VEGF-Induced Adult Neovascularization:
Recruitment, Retention, and Role
of Accessory Cells
Myriam Grunewald,1,4Inbal Avraham,1,4Yuval Dor,1Esther Bachar-Lustig,2Ahuva Itin,1Steffen Yung,2
Stephano Chimenti,3Limor Landsman,2Rinat Abramovitch,1and Eli Keshet1,*
1Department of Molecular Biology, Hebrew University-Hadassah Medical School, Jerusalem 91120, Israel
2Department of Immunology, The Weizmann Institute of Science, Rehovot 76100, Israel
3The Mario Negri Institute for Pharmacological Research, Milan 20157, Italy
4These authors contributed equally to this work.
Adult neovascularization relies on the re-
cruitment of circulating cells, but their an-
are unclear. We show that the endothelial
growth factor VEGF is sufficient for organ
homing of circulating mononuclear myeloid
cells and is required for their perivascular
positioning and retention. Recruited bone
marrow-derived circulating cells (RBCCs)
summoned by VEGF serve a function dis-
tinct from endothelial progenitor cells. Re-
tention of RBCCs in close proximity to
angiogenic vessels is mediated by SDF1,
a chemokine induced by VEGF in activated
perivascular myofibroblasts. RBCCs en-
via secreting proangiogenic activities dis-
tinct from locally induced activities. Pre-
cluding RBCCs strongly attenuated the
proangiogenic response to VEGF and addi-
esis in excision wounds. Together, the data
suggest a model for VEGF-programmed
adult neovascularization highlighting the
Formation of new blood vessels in adult organs is funda-
mentally different from developmental neovascularization.
Whereas developmental vasculogenesis relies on a reservoir
of local angioblasts, adult neovascularization may utilize cir-
culating endothelial progenitor cells (EPCs) mobilized from
the bone marrow (BM) and eventually incorporated within
the forming vasculature at a distant organ (Asahara et al.,
1999; Asahara et al., 1997; Lyden et al., 2001; Rafii and
Lyden, 2003). The relative contribution of circulating EPCs
able, and may range from a minor (Machein et al., 2003;
Peters et al., 2005; Rajantie et al., 2004; Wagers et al.,
2002; Ziegelhoeffer et al., 2004) to a major contribution
(Garcia-Barros et al., 2003; Hattori et al., 2001), presumably
reflecting also differences in the genetic background, the
organ (or tumor) involved (Ruzinova et al., 2003), and the na-
ture of the angiogenic stimulus.
Recently, increased attention has been directed to other
populations of BM-derived cells recruited to sites of ongoing
angiogenesis but that do not function as EPCs. These cells
might nevertheless contribute to neovessel formation (De
Palma et al., 2003; Takakura et al., 2000) by a yet-unknown
mechanism. The present study focuses on the angiogenic
roles of these cells, collectively defined here as recruited
bone marrow-derived circulating cells (RBCCs) and on their
significance in light of the need to reconcile the apparent lim-
ited contribution of EPCs with the apparent beneficial effect
of autologous BM cells administered to ischemic tissues
(Kalka et al., 2000; Kocher etal., 2001), findings that have al-
ready prompted clinical trials (Schachinger et al., 2004). In
addition, understanding the unknown mechanisms underly-
menting or, conversely, attenuating angiogenesis.
Another unique aspect of adult neovascularization is that,
unlike the hardwired program of developmental angiogene-
sis, it is usually triggered by either environmental stress (e.g.,
hypoxia) or by stochastic genetic changes in tumors. These
of VEGF. It is, therefore, mechanistically unclear how VEGF
single-handedly orchestrates a complete angiogenic pro-
gram, a task known to be executed by multiple and often
Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc. 175
nonredundant factors. For example, the fact that VEGF is
a relatively weak endothelial cell mitogen suggests that addi-
tional EC mitogens are likely to be recruited. The present
plementary proangiogenic activities, not only through local
anism must exist by which VEGF acts to recruit accessory
cells and position them close to angiogenic vessels within
the target organ.
To address these issues, we devised a transgenic system
for conditional and reversible induction of VEGF in selected
adult organs. The system has several advantages over pre-
viouslyemployed systemsofVEGF-induced neovasculariza-
tion in the adult. First, as a ‘‘clean’’ genetic switch system,
it circumvents confounding factors associated with other
modes of VEGF delivery, such as stress, inflammatory, and
functions in the absence of hypoxia, an environmental factor
recently shown to influence trafficking of hematopoietic cells
(Ceradini et al., 2004). Second, the synchronous and robust
reaction to VEGF enables to summon onto an organ of
choice circulating cells in quantities allowing their isolation
and subsequent characterization. Third, the possibility to
switch off VEGF production at any desired schedule allows
cells from a role in maintaining incoming cells in the organ.
Here, we provide evidence that VEGF induces perivascu-
lar expression of the chemokine SDF1 that functions, in turn,
to position RBCCs in this strategic location from which they
act in a paracrine fashion to enhance in situ proliferation of
resident, activated endothelial cells.
A Genetic System for Conditional and Reversible
Induction of VEGF in Selected Adult Organs
To induce formation of new blood vessels in selected adult
organs, a transgenic system in which VEGF expression is in-
duced at will in the respective organ, steadily maintained for
the desired duration, and subsequently can be switched off
was designed. Briefly, transgenic mice expressing a tetracy-
cline-regulated transactivator protein (tTA) exclusively in the
myocardium or liver (driver lines) were mated with transgenic
mice harboring a VEGF164-encoding transgene driven by
a tetracycline-responsive promoter (responder line). Pups
that inherited both transgenes were selected for modulating
VEGF expression, whereas littermates that inherited only
one of the two transgenes served as controls. The onset of
VEGF induction in these animals and the duration of expres-
sion were tightly controlled by including or omitting tetracycline
from the drinking water (‘‘off’’ and ‘‘on’’ modes, respectively;
see Experimental Procedures for details). Previous studies
from our laboratory indicated that in MHCa-tTA+/tet-VEGF+
animals, VEGF164 expression was strongly induced within
24 hrafterwithdrawal oftetracycline in mostcardiomyocytes
and in PLAP-tTA+/tet-VEGF+animals in most hepatocytes
(Dor et al., 2002). Ongoing production of VEGF by the liver
resulted in its accumulation in the serum to steady-state
levels of 120 to 3000 pg/ml and when induced in the heart,
to levels of 70 to 140 pg/ml. For comparison, control mice
had only 45–100 pg/ml VEGF in their circulation. The appar-
to determine the threshold levels of circulating VEGF condu-
cive for an ensuing angiogenic response. Using BrdU+prolif-
erating endothelial cells as a readout for angiogenesis, a sig-
nificant angiogenic response was already detected for the
lower end of this range in the liver (see Figure S1 in the Sup-
plemental Data available with this article online) as well as in
the heart. Notably, these levels of circulating VEGF are com-
parable and even lower than the respective levels detected
in cancer patients (Kraft et al., 1999; Salven et al., 1997), in
neovascularization associated with acute myocardial infarc-
tion (Ogawa et al., 2000), or during wound healing (Infanger
et al., 2004). Thus, we believe that findings obtained with
the aid of this transgenic switch system are also valid for nat-
urally triggered angiogenesis, as indeed extended below for
VEGF Induces Homing of Circulating Myeloid Cells
into the Respective Target Organ
Inducing VEGF expression in the adult heart or liver led to
massive infiltration of circulating cells, specifically into the
respective organ. Recruited cells could be readily distin-
guished morphologically from the resident cells of the re-
spective organ (Figure 1A). Infiltration of circulating cells was
already evident shortly after the onset of VEGF induction,
clearly visible within the first 4 days, and notably, preceding
emergence of new vessels. Thus the notion that these re-
cruited cells maycontribute to neovessel formation isatleast
To track the BM origin of recruited cells, the marrow of
double-transgenic animals was reconstituted with geneti-
cally tagged cells prior to switching on VEGF expression.
Specifically, BM cells derived from ROSA-bGal mice or, in
other experiments, marrow derived from b-actin-eGFP mice
were used to replace the marrow of irradiated double trans-
genic recipients (see Experimental Procedures for details
and for analysis of chimerism). In the noninduced liver, only
very few bGal- or eGFP-labeled cells were detected (data
not shown), whereas following VEGF induction, the organ
was heavily populated by BM-derived cells (Figures 1B and
1C), indicating that their accumulation in the liver is due to
active recruitment by VEGF and not merely reflecting a natu-
ral turnover of organ resident BM-derived cells that have
taken place since BM grafting.
Recruited RBCCs summoned to the organ by VEGF were
only rarely incorporated within the endothelium, arguing
against a significant contribution of EPCs to the newly ac-
quired vasculature in this system (Figures 1B and 1C). In-
stead, RBCCs were mostly distributed around blood vessels
spatial relationship between RBCCs and endothelial cells
suggested a paracrine angiogenic role for RBCCs rather
than a role of serving as endothelial progenitors.
176 Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc.
Figure 1. VEGF Recruits Circulating Hematopoietic Cells to the Organ from which It Emanates
VEGF expression was induced in the heart or liver of newborn or 4 weeks old mice, respectively, as described in the Experimental Procedures, and the
respective organ was retrieved 2 weeks thereafter.
(A) H&E-stained sections of the myocardium and liver maintained in the VEGF ‘‘off’’ (left) or ‘‘on’’ (middle) modes. Right images are from mice in which VEGF
expression was switched on for 2 weeks (at this time, mice had a level of circulating VEGF comparable to the mice shown in the middle lanes) and sub-
sequently switched off for 2 weeks. Note the presence of recruited cells in the organ maintained in the ‘‘on’’ mode and their complete loss following
(B and C) The bone marrow of PLAP-tTA/VEGF+animals was replaced with genetically tagged marrow donated by either ROSA-bGal mice (B) or b-actin-
EGFP transgenic mice (C) described in the Experimental Procedures. Following a complete BM reconstitution, VEGF expression was switched on for
2 weeks. Ten micrometer-thick liver cryosections were stained for b-galactosidase activity or viewed for GFP fluorescence (in sections also stained for
vWF), respectively. Note that BM-derived cells are clustered around blood vessels. A higher magnification of the boxed area in (B) clearly shows that re-
cruited cells assume a periendothelial location but that BM-derived cells are not incorporated within the endothelium (which is indicated by arrowheads).
Likewise, in (C), GFP+cells reside outside the vWF+endothelium.
Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc. 177
VEGF is known to induce mobilization of BM cells into the
bloodstream, as reflected in a significant increase in WBC
counts by circulating VEGF (Hattori et al., 2001). In our sys-
tem, an 8-fold increase in WBC count was observed when
VEGF was induced in the liver, but only a minor increase in
WBC count was observed following a heart switch. This is
likely because the levels of circulating VEGF resulting from a
heart switch were significantly lower. Yet, even in the case
where mobilization of BM cells by VEGF was not excessive,
circulating cells were efficiently attracted to and accumu-
lated within the target organ. Thus, we consider the role of
VEGF in homing circulating cells to the target organ to be
of a greater significance than its role in mobilizing cells into
the circulation, a task known to be accomplished also by
other cytokines (like GM-CSF).
To elucidate the cellular composition of RBCCs, we de-
vised a procedure for their isolation in entirety after their ar-
rival to the liver or heart. Two weeks after the onset of VEGF
mulated in the respective target organ to allow their isolation
in sufficient quantities. Importantly, at this time there was no
evidence for tissue inflammation or damage that might be
associated with a prolonged exposure to VEGF. Advantage
was then taken of the fact that recruited cells are loosely
associated with the liver for their separation, without the
need for proteolysis and for their further purification through
banding on a Ficoll gradient. The yield of liver RBCCs was
1.5 ? 107cells per organ whereas only 1% to 10% that
amount could be isolated from the noninduced liver using
the same procedure. FACS analyses have shown that the
vast majority of isolated cells are positive for the panhemato-
poietic marker CD45 (Figure 2A, left), positive for VEGF-R1
and negative for VEGF-R2 (data not shown), indicating that
cells recruited by VEGF are predominantly hematopoietic in
nature. In situ hybridization with a CD45-specific riboprobe
corroborated the observation that RBCCs are retained close
to the vessels from which they have apparently extravasated
showed that 40% of CD45+RBCCs retrieved from the liver
were also CD11b positive (data not shown).
Since VEGF can induce extra-medullary hematopoiesis in
the liver (Hattori et al., 2001; our unpublished data) and,
therefore, the retrieved cell population could have also con-
tained cells engaged in active hematopoiesis, we extended
a similar analysis to RBCCs retrieved from induced hearts.
Like in the liver, 66% of CD45+cell recruited by VEGF to
the heart were also CD11b positive (Figure 2A, right). Nota-
bly, B and T lymphocytes could not be detected in heart
RBCCs, as evident by the lack of CD19 or CD3 expression
(data not shown), suggesting that the major components in
the RBCC population are myeloid cells.
To show that the perivascular positioning of RBCCs dem-
onstrated above for CD45 cells, in general, specifically ap-
plies to myeloid cells, we bred onto the VEGF switch system
a transgenic GFP reporter which is driven by the CX3CR1
promoter. Activity of this chemokine receptor promoter is re-
stricted to mononuclear myeloid cells, including all circulat-
ing CD116+monocytes, and is essentially absent from cells
of the lymphoid lineage (Geissmann et al., 2003; Jung et al.,
2000). As shown in Figure 2C, CX3CR1/GFP-positive cells
were indeed localized close to vessels in both the liver and
heart. This localization has suggested that recruited mono-
nuclear myeloid cells may act paracrinically on the adjacent
endothelium and prompted experiments described below.
VEGF Induces Perivascular Expression of the
Chemokine SDF1 that Is Responsible for Positioning
RBCCs Close to Blood Vessels
To explain how VEGF alone may act to retain RBCCs in a
perivascular position, we hypothesized that VEGF induces
expression of a particular chemokine that functions to cap-
ture incoming cells expressing cognate receptors. As a first
step, we determined the repertoire of chemokine receptors
expressed in incoming cells by analyzing RNA extracted
from isolated RBCCs on an Affymetrix gene array. Out of
the several C-C and C-X-C type chemokine receptors ex-
pressed by these cells, we focused on CXCR4 because it
was found to be expressed by the vast majority of CD45+
cells that have entered the organ (Figure 3A) and since its
obligatory ligand CXCL12 (also known as SDF1) was previ-
ously found to be important for recruiting hematopoietic
stem cells to the liver (Kollet et al., 2003). We first verified
that incoming RBCCs are indeed responsive to SDF1 using
a transfilter migration assay. As shown in Figure 3B, RBCCs
retrieved from the liver showed a strong chemotactic re-
sponse to SDF1.
Next, we examined whether SDF1 is naturally induced
downstream of VEGF and in what locations. Immunohisto-
chemical analysis showed that switching on VEGF expres-
sion in either heart or liver resulted in induction of SDF1 pro-
tein, predominantly around blood vessels of the respective
organ (Figure 3C) and to a lesser extent also in the endothe-
lium (data not shown). Further, in situ mRNA hybridization for
SDF1 indicated that perivascular cells are the producers of
this secreted chemokine in the VEGF-induced organ (Fig-
serial sections confirmed an intimate spatial relationship
where incoming CXCR4+cells are retained as a halo around
the ring of SDF1-expressing cells surrounding vessels
To better identify the SDF1-expressing cells, double
immunostaining was performed using SDF1 antibodies in
combination with fibroblastic or smooth muscle/pericyte
markers. In the VEGF-induced heart, SDF1 protein mostly
colocalized with a-smooth muscle actin (aSMA; Figure 4A,
left panels) and with vimentin (data not shown), suggesting
the fibroblastic or smooth muscle nature of SDF1-express-
Next, we wished to demonstrate that perivascular expres-
sion of SDF1 also accompanies pathophysiological neovas-
cularization. To this end, specimens representing ongoing
angiogenesis in the ischemic heart (experimentally induced
by ligation of the left anterior descending coronary artery
[LAD]), in a growing prostate tumor and during healing of ex-
cision wounds, were similarly analyzed for SDF1 and aSMA
178 Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc.
expression. As shown in Figure 4A, colocalized expression
of SDF1 and aSMA in perivascular cells was indeed ob-
served in all cases, indicating that perivascular induction of
SDF1 is an integral part of adult angiogenesis in general.
As evident in Figure 4A, with the exception of the VEGF ge-
netic switch system, SDF1 expression was detected in addi-
tional sites besides its perivascular expression. This was an-
ticipated, considering that SDF1 is known to be regulated by
different cues (e.g., by hypoxia [Ceradini et al., 2004]). It was
iological settings too, SDF1 is regulated by VEGF. To this
end, VEGF signaling was specifically inhibited during wound
Figure 2. RBCCs, Mostly Myeloid Cells, Are Retained in a Perivascular Position
(A)(Left)Cells recruited tothe liverfollowing a2week VEGF switch wereisolatedasdescribedinthetextand analyzedbyFACSforCD45expression.(Right)
Heart RBCCs were purified and sorted for CD45 and CD11b expression. The figure depicts CD45+-gated cells showing that about 2/3 of them are also
(B) In situ mRNA hybridization with a CD45-specific probe of the VEGF induced liver showing recruited CD45+surrounding a blood vessel. Left and right
figures are brightfield and darkfield images, respectively.
(C) CX3CR1-GFP transgenic mice were bred with double-transgenic mice harboring either the liver- or heart specific-switchable system. VEGF expression
was induced in triple-transgenic mice as described in Figure 1, and CX3CR1/GFP-expressing cells before or after inducing VEGF were visualized by GFP
fluorescence. Note the perivascular positioning of recruited myeloid cells in both the induced liver and heart.
Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc. 179
healing angiogenesis through transgenic induction of a chi-
meric protein that acts as a ‘‘VEGF trap’’ (see Experimental
Procedures for details). Specimens were then retrieved at
the midst of the wound healing process and were analyzed
upon wounding the skin, but its upregulated expression was
negated when the VEGF trap was activated (Figure 4B and
a quantitative summary in Figure 7A). Importantly, inability
to upregulate SDF1 in the wound in the face of VEGF sup-
pression was specifically accounted for by the apparent inhi-
bition of its perivascular expression (Figure 4C), suggesting
that in wound angiogenesis, induction of perivascular SDF1
is downstream of VEGF.
Currently, it is not known whether induction of SDF1 ex-
pression in these cells by VEGF is direct or indirect. In vitro,
at least, VEGF directly induced SDF1 mRNA in primary fibro-
blast cultures, but not in primary cultures of smooth muscle
cells (Figure S2).
Ongoing Stimulation of the VEGF-SDF1 Pathway Is
Essential for Maintaining Imported RBCCs
within the Organ
Results presented above indicated that VEGF recruits circu-
lating cells to the organ where it is produced. We next exam-
ined whether a continued VEGF expression is required for
retaining imported cells within the organ and also whether
RBCC retention is mediated by SDF1. To this end, we in-
duced VEGF expression for two weeks and then switched
it off. While the liver or the heart contained an abundance
of RBCCs by the end of the induction period, recruited cells
were completely lost within a few days after terminating
VEGF expression (Figure 1A). The loss of recruited RBCCs
was also evident from a 10- to 100-fold reduction in the yield
of RBCCs that could be isolated from livers in which VEGF
expression has been switched off. In the organ in which
VEGF has been switched off, physical loss of RBCCs was
also manifested by a marked reduction in organ CXCR4
Responsive to SDF1 Induced by VEGF in
(A) FACS analysis of RBCCs retrieved from the
liver, showing that the majority of CD45+cells
are also positive for CXCR4.
(B) 2 ? 105RBCCs retrieved from the liver were
placed in the upper chamber of a transfilter mi-
gration device. Cells migrated toward a serum-
free control medium or toward a serum-free me-
dium containing 250 ng/ml SDF1 protein were
counted by FACS (number of cells/30 s).
(C) Immunostaining for SDF1 in the noninduced
(off) and VEGF-induced (on) heart and liver. Note
that VEGF induces SDF1, specifically in perivas-
constitutively expressed in the bile duct epithe-
(D) In situ hybridization of SDF1a and CXCR4
mRNAs. Two serial sections of a VEGF-induced
liver were hybridized with a SDF1- or a CXCR4-
specific antisense riboprobe. The respective hy-
bridization signals, visualized by darkfield illumi-
nation and pseudocolored in red (SDF1) or blue
(CXCR4), were superimposed to produce the im-
age on the right. Note that extravasated CXCR4+
cells (clearly visible in the brightfield image on the
left) are distributed as a halo around SDF1-
expressing perivascular cells. Arrow points at the
bile duct epithelium expressing SDF1 mRNA.
180 Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc.
expression, the presumed consequence of the observed
downregulation in SDF1 expression (Figure S3A). A role for
SDF1 in RBCCs retention during wound healing was sug-
gested by a similar coreduction in SDF1 and CXCR4 expres-
sion in the wound area upon activation of the VEGF trap (Fig-
ure 4B, quantified in Figure 7A).
and/or retention of incoming CXCR4+cells, we inhibited
used as a CXCR4-specific inhibitor (reviewed in De Clercq
). In contrast to VEGF-induced animals harboring a
present, CXCR4+-expressing cells could not be detected in
the livers of VEGF-induced animals harboring an osmotic
pump releasing AMD3100 (Figure S3B). Together, the data
support a scenario where SDF1 functions downstream of
VEGF to maintain these cells within the target organ.
Ectopic expression of SDF1 in the noninduced liver,
achieved via adenovirus-mediated delivery of SDF1 cDNA,
ectopically expressed SDF1 partially prevented the loss of
RBCCs already recruited to the liver by VEGF, a loss that
would have otherwise taken place upon VEGF withdrawal
(Figure S4). These findings suggest that SDF1 acts primarily
for these cells (see Discussion).
RBCCs Are Required for Angiogenesis via
Enhancing In Situ Proliferation of
We next wished to determine whether VEGF-instructed re-
cruitment of RBCCs is an integral component of adult organ
neovascularization and, specifically, to examine whether
these imported cells are essential for efficient neovasculari-
zation. First, the potential of RBCCs to promote sprouting
angiogenesis ex vivo was examined. When placed atop col-
lagen-embedded mouse aortic segments, RBCCs isolated
from the induced liver induced extensive sprouting at a mag-
nitude that even exceeded that induced by a recombinant
VEGF protein (Figure 5A and quantification in Figure 5B).
To show that this is not a peculiarity of cells isolated from
the liver, RBCCs summoned by VEGF to the heart and sim-
moted vessel sprouting at a comparable efficiency to that of
liver RBCCs (Figure 5B). These experiments were repeated
(data not shown). Further, a strong sprouting-promoting ac-
tivity was attributed to the CD11b+cells purified to homo-
geneity from the total RBCC population through capturing
on magnetic beads conjugated to anti-CD11b antibodies
(Figure 5B). Medium conditioned by each of these cell pop-
ulations was as efficient as the respective cells in supporting
sprouting (Figure 5), suggesting that vessel wall endothelium
is likely activated by proteins secreted by nearby RBCCs in
a paracrine fashion.
The repertoire of proangiogenic activities elaborated by
RBCCs remains to be determined. As a first step, the profile
of RBCC-expressed genes was elucidated through subject-
ing RNA isolated from purified RBCCs to a gene-array anal-
ysis. Candidate genes known to encode secreted proteins
are then tested with respect to the ability of their specific in-
hibitors to block RBCC-initiated sprouting. As an example,
MMP9 is highly expressed by RBCCs and its blockade, us-
ing the specific inhibitor GM6001, strongly inhibited RBCCs-
driven sprouting angiogenesis ex vivo (Figure 5B). Impor-
tantly, the predominant contribution of MMP9 is by RBCCs
and not by cells that are natural inhabitants in the VEGF-
induced organ (comparative Affymetrix array data, not
shown [see Discussion]).
To provide a definite in vivo proof that RBCCs recruited
by SDF1 contribute to organ neovascularization, cells were
prevented from reaching and/or staying within the organ
through implanting an AMD3100-releasing pump prior to
switching on VEGF, and the consequences on organ neo-
vascularization were determined. Briefly, hepatic expression
of VEGF was induced in PLAP-tTA+/VEGF+littermates, half of
which were implanted with an AMD3100-releasing pump
and the other half with a saline-releasing pump. Fourteen
days later, livers were retrieved and analyzed with respect
to the following parameters: the level of transgenic VEGF in-
duced in the organ, the content of CXCR4+cells populating
the organ, and the magnitude of the angiogenic response.
Blocking SDF1/CXCR4 interaction led to a significant inhibi-
tion of liver neovascularization (Figure 6). A pairwise compar-
ison of two representative littermates is shown in Figure 6A.
denced by a comparable level of transgenic VEGF induced
in the organ. Yet, the apparent failure to recruit CXCR4+cells
inthe AMD3100-treatedanimal wasassociated withamark-
edly reduced angiogenic response, as already suggested by
a significant inhibition in VE-cadherin expression (Figure 6A).
A compromised angiogenic response was clearly evident
from comparing the elaborate network of blood-filled neo-
vessels induced close to the surface of the control liver to
the scarcity of induced vessels in the AMD3100-treated
animal (Figure 6B). Inspection of histological sections con-
firmed a significant reduction in the number of large lume-
nized and sinusoidal structures observed in sections of
the AMD3100-treated animals (Figure 6B). Data compiled
for several litters (excluding only animals in which the level
of induced circulating VEGF did not reach the set threshold)
were quantified (Figure 6C). The control and AMD3100-
treated groups were very similar with respect to the level of
transgenic VEGF induced, but the treated group had on av-
erage only one-fifth the normal amount of CXCR4+cells re-
cruited to the organ. Inhibition of neovascularization in the
AMD3100-treated group was again evident from visual in-
spection ofwholemounts andliversections andwasalsore-
flected in a 2.5-fold reduction in VE-cadherin expression. It
should be pointed out, however, that quantifying neovascu-
larization in the liver through marker analysis is difficult due to
the fact that in this highly vascularized organ, newly added
vessels comprise a relatively small fraction of vessels. More-
over, some of the markers routinely used to mark vessels
(e.g., CD31) are expressed on RBCCs. Therefore, to directly
focus on the angiogenic endothelium, we resorted to direct
Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc. 181
Figure 4. SDF1 Induction in aSMA+Perivascular Cells during Pathophysiological Angiogenesis Is VEGF Dependent
(A) Double immunostaining for aSMA (red) and SDF1 (green) in the following settings of angiogenesis: (1) myocardial angiogenesis induced by a genetic
VEGF switch, (2) in the ischemic myocardium following experimental occlusion of the LAD coronary artery, (3) in a 22RV1 prostate carcinoma grafted
into a SCID mouse, (4) in a skin excision wound 48 hr postwounding.
(B and C) Wounded skin tissue was removed at 48 hr postwounding and divided symmetrically in two halves. One was used for extracting RNA (B) and the
other for preparing tissue sections (C).
182 Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc.
measurements of endothelial cell proliferation. BrdU was in-
jected into mice with an ongoing heart or liver VEGF switch,
cells were then visualized by BrdU immunohistochemistry.
In both organs, extensive endothelial cell proliferation took
place almost exclusively in endothelial cells residing in the
vessel wall (Figure S5). To examine whether CXCR4 inhibi-
tion will negatively impact on proliferation of vessel-wall en-
dothelium, BrdU pulse experiments were also performed in
the presence of AMD3100. CXCR4 inhibition and the resul-
tant failure to retain recruited cells led to a 3-fold reduction in
the number of vessels that contained one or more BrdU-
positive cells and nearly 8-fold reduction in the total number
of proliferating endothelial cells (Figure 6C). These results
indicate that RBCCs are required for efficient proliferation
of vessel wall endothelium.
To substantiate that the VEGF-SDF1-RBCCs assisted an-
giogenesis connection also operates during wound healing
angiogenesis, the natural process was manipulated in three
different ways. Namely, inhibition of VEGF signaling through
the use of a VEGF trap (some representative results were
already presented in Figure 4), blocking SDF1 signaling
through the use of AMD3100, and administration of isolated
RBCCs. As summarized in Figure 7A, VEGF suppression
abrogated upregulated expression of SDF1 in the wound,
inhibited CXCR4 expression, and completely blocked angio-
genesis (here, VE-cadherin was used as a vascular readout).
SDF1 blockade by AMD3100 similarly inhibited CXCR4 and
angiogenesis. Conversely, injection of purified RBCCs to
the wound area significantly augmented neovascularization.
These results indicate that RBCCs recruited by VEGF via
SDF1 are instrumental also to wound healing angiogenesis.
This study focuses on the poorly understood function of cir-
culating cells recruited to sitesofongoing neovascularization
in the adult. While infiltration of circulating cells to angiogenic
sites has been observed frequently, the notion that BM-de-
rived cells are actually required for adult neovascularization
(De Palma et al., 2003) needed to be substantiated. We
use the term RBCCs to include all types of recruited cells
and focused on the function of RBCCs that are not EPCs.
Here, we resorted to a transgenic system designed for in-
inflammation, and tissue injury in order to provoke recruit-
ment of circulating cells. With the aid of this system, we
showed that VEGF alone is sufficient to instruct homing of
hematopoietic cells onto the organ from which it emanates.
Together with evidence that these cells play an indispens-
able role in neovessel formation, findings reported here es-
tablish that VEGF-induced recruitment of RBCCs is an inte-
gral component of adult neovascularization.
vasculatures is currently debated and appears to greatly
depend on the experimental system studied. In our system,
representing adult organ neovascularization, the contribu-
tion of EPCs to neovessels was very low. Results of the bone
marrow transplantation experiments (Figures 1B and 1C)
were consistent with results of the BrdU pulse labeling ex-
periments (Figure S5), indicating that the major source for
new endothelial cells istheirexpansion via in situproliferation
of vessel-wall endothelium. It could be argued, nevertheless,
that BrdU-positive endothelial cells may represent preinte-
grated EPCs with a high proliferation capacity. This was not
compatible, however, with the failure to detect strings of
bGal+or GFP+endothelial cells (i.e., ECs of a BM origin) in
the neovascularized organ.
The precise composition of the RBCCs population, pre-
sumed to be heterogeneous, was not fully determined. Par-
ticularly, marker combinations typifying perspective EPCs
contribution. Quantitatively, analysis of the entire RBCCs
population, retrieved from two different organs, has clearly
component. The current confusion regarding the nature of
circulating cells recruited to angiogenic sites results in part
from a lack of clear criteria for their isolation from the blood-
stream. For example, cells isolated from the human circula-
tion and defined by some investigators as EPCs were shown
by others as mostly nonproliferating cells expressing mono-
on the other hand, has used the target organ as a functional
filter to isolate the angiogenically relevant cells, thereby cir-
cumventing the need to rely on marker expression.
by virtue of their VEGF-R1 expression (Barleon et al., 1996).
dantly corecruited with EPCs to angiogenic tumors (Lyden
et al., 2001). It appears that RBCCs are mostly naturally
circulating mononuclear myeloid cells endowed with the
capacity to secrete proangiogenic activities (Figure 5). This
argues that RBCCs’ capture and positioning might be of
a greater significance than their BM mobilization.
The paracrine mode of RBCCs action necessitates a
mechanism to retain incoming cells nearby the endothelium
upon which they act. Findings reported here disclosed
a mechanism by which RBCCs are retained close to blood
pression that functions to capture incoming CXCR4+cells.
Extending the observation of perivascular SDF1 induction to
the major settings of natural angiogenesis and, in particular,
(B) Northern blot analysis for SDF1 and CXCR4 mRNAs in the intact skin, wounded skin, and wounded skin of mice in the presence of a chimeric soluble
VEGF-R1 protein (VEGF trap). The soluble VEGF-R1 protein was induced 7 days before wounding and was present in the circulation during the healing
process at concentrations of 2062 pg/ml or 1150 pg/ml (for the respective left and right lanes).
Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc. 183
demonstrating that perivascular SDF1 induction during
wound healing is VEGF dependent (Figure 4), indicated
that SDF1 upregulation is an integral part of a general pro-
gram set in motion by VEGF to localize RBCCs close to an-
giogenic vessels (see model in Figure 7B).
Experiments reported here show that ongoing VEGF sig-
naling is essential to maintain incoming CXCR4+cells within
the target organ. Since SDF1 is also a chemoattractant for
these cells (Figure 3) it is plausible that SDF1 operates in
both homing of CXCR4-expressing cells as well as for their
retention within the organ. Alternatively, homing might be
mediated by other factors, like VEGF, a possibility supported
by findings that blocking VEGF-R1 reduces the number of
recruited perivascular cells in tumors (Hattori et al., 2001).
Figure 5. RBCCs Retrieved from VEGF-Induced Heart and Liver Promote Sprouting Angiogenesis Ex Vivo
Mouse aortic segments were embedded in collagen and overlaid with either a serum-free control medium or a serum-free medium containing 10 ng/ml
VEGF, or 6 ? 104RBCCs retrieved from the indicated organ with or without further fractionation, or overlaid with 500 ml serum-free medium conditioned
was measured; a relative sprouting index was calculated as described in the Experimental Procedures and is presented relative to VEGF-induced sprouting
(B). The MMP inhibitor GM6001 was added at a concentration of 25 mM.
184 Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc.
not explain the perivascular localization of recruited cells,
considering the uniform distribution of VEGF throughout
the organ. Another possibility is that RBCCs passively ex-
travasate the VEGF-activated endothelium and thus the
main requirement is for SDF1-mediated retention. Thus, it
Figure 6. SDF1 Blockade Inhibits VEGF-Induced Neovascularization via Reducing In Situ Proliferation of Endothelial Cells
Mice were implanted with a control, saline-releasing osmotic pump or with an AMD3100-releasing osmotic pump at the onset of inducingVEGF expression
in the liver. Two weeks later, the liver was excised and analyzed as indicated.
(A) A pairwise comparison of two representative littermates. Top left: Semiquantitative RT-PCR using primers detecting transgenic (but not endogenous)
VEGF.Noteacomparable level ofVEGF inducedinmiceimplantedwithacontrolpump(?A)oranAMD3100pump(+A).Topright:NorthernblotforCXCR4
and VE cadherin.
(B)Left:Top view ofthe livershowing adensenetworkofsuperficial vessels inducedbyVEGF inthe controlanimal that isstrongly inhibited intheAMD3100-
treated animal. Right: H&E-stained sections of the same liver lobe shown above showing that most of the deeper plexus of VEGF-induced neovessels are
inhibited by AMD3100. The large channels seen in the right image are sections through the normal portal triad (PT).
(C) Data compiled from nine mice implanted with a controlpump and 11 mice implanted withan AMD3100 pump (comprising all double-transgenic animals
inthree consecutive litters inwhich a successful VEGF switch was evidenced by >200 pg/mlof circulating VEGF). Animals were implanted with eithera con-
trol pump or an AMD3100-releasing pump at the time of switching on hepatic VEGF expression. Quantification of transgenic VEGF was by semiquantitative
PCR normalized to L19 RNA and of CXCR4 and VE-cadherin by densitometry of Northern blots. For measurements of BrdU+cells, BrdU was injected 3 hr
63 vessels, respectively (identified as cells lining the lumen of erythrocyte-containing structures).
Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc. 185
is possible that RBCCs in our liver system extravasate
through sinusoidal endothelial cells but are retained mostly
around large veins. The action of SDF1 as a RBCCs retainer
is also consistent with our finding that ectopic SDF1 expres-
sion partially prevents RBCC loss upon VEGF withdrawal
(Figure S3) and with findings that the apparent increase in
angiogenesis by exogenous SDF1 is only manifested in the
presence of another insult like hypoxia (Abbott et al., 2004).
Modulations in SDF1 expression may greatly effect traf-
ficking of hematopoietic cells into and away from the bone
marrow (Peled et al., 1999) and to specific microdomains
ingCXCR4+metastatic cells toorgans thatproduce it(Muller
et al., 2001) and to control trafficking of circulating cells
to hypoxic sites (Ceradini et al., 2004). A recent study has
shown that recruitment of EPCs to breast carcinomas is me-
diated in part by SDF1 secreted by carcinoma-associated
fibroblasts (Orimo et al., 2005). Ours is the first study, how-
ever, to show that SDF1 functions downstream of VEGF to
Figure 7. The VEGF-SDF1-RBCC-Assisted Angiogenesis Connection Is Operative during Wound Healing
(A) Skin wounds were performed as described above and were manipulated as follows: without further manipulation (n = 5); A VEGF trap was activated as
described in Figure 4B (n = 4); for SDF1 blockade, an osmotic pump releasing AMD3100 was implanted prior to wounding as described above (n = 4); for
RBCCs administration, 1.105cells retrieved from a VEGF-induced liver were suspended in 100 ml PBS and injected s.c. close to the wound, at 30 min post-
wounding (n = 9). RNAs were extracted from tissues retrieved at 48 hr after wounding and were subjected to a Northern blot analysis with the indicated
probes. Hybridization signal intensities were measured and compared with those obtained with intact skin RNA (cohybridized on the same blot) that
was given a value of 1.
(B) A model for a program induced by VEGF for the recruitment and paracrine localization of angiogenic accessory cells.
186 Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc.
Findings reported here highlight a novel principle in
VEGF-induced neovascularization, namely, a dual source
of proangiogenic activities that complement VEGF. That is,
‘‘imported’’ paracrine activities contributed by RBCCs and
activities induced by VEGF locally. Ongoing experiments at-
tempt categorizing proangiogenic activities according to
their respective source. In these experiments, the transcrip-
at the ‘‘off’’ and ‘‘on’’ modes. An example for a RBCCs-
contributed activity is MMP9, which has been shown to be
an important component of angiogenesis (Bergers et al.,
2000; Figure 5B). Examples for genes upregulated solely in
the organ resident cells are multiple components of the
PDGF and Notch pathways (data not shown). It can be fur-
ther speculated that RBCCs are reprogrammed by VEGF
and/or the organ milieu for more efficient secretion of proan-
giogenicproteins.Thisintriguingpossibility iscurrently inves-
tigated through comparing expression profiles of CD11b+
cells isolated from the circulation with that of CD11b+cells
retrieved from the heart of the same VEGF-induced animal.
Finally, experiments reported here showing that inhibition
of SDF1 impairs organ neovascularization suggest that the
SDF1/CXCR4 system might be harnessed as a new target
for antiangiogenesis. Conversely, the findings provide
a mechanistic rationale for using SDF1 (Yamaguchi et al.,
2003) or cells engineered to express CXCR4+to augment
Transgenic Mice and Conditional Modulations of
VEGF Expression and Signaling
Heart-specific induction was achieved using a transgenic driver line in
which tTA expression is driven by a myosin heavy chain-a heart specific
promoter (MHCa; Yu et al., 1996). Liver-specific induction was obtained
by using a driver line in which tTA expression is driven by a C/EBPb
(CCAAT/enhancer binding protein b) promoter (also known as liver-
activator protein or PLAP; Kistner et al., 1996). The responder tet-
VEGF164 transgenic line was previously described (Dor et al., 2002).
The responder tet-VEGF TRAP transgenic line encodes a tetracycline-
inducible protein composed of an IgG1-Fc tail fused to the extracellular
domain of VEGF-R1 (corresponding to amino acid residues 1 to 631 in
Induction of VEGF or of a soluble VEGF-R1 (‘‘VEGF trap’’) by tetracy-
cline withdrawal and its shut-off by tetracycline addition were carried out
as previously described (Dor et al., 2002). VEGF protein levels in sera of
induced mice were measured using mouse VEGF ELISA kit (Oncogene
Inc.),andlevels ofcirculatingVEGFtrapproteinweredetermined byasol-
uble VEGF-R1 ELISA kit (R&D Systems).
Isolation and Analysis of Recruited Cells
Circulating cells recruited to the liver of 8- to 10-week-old mice following
VEGF induction were isolated by filtrating through 100 Mesh (Sigma) and
purified by density gradient centrifugation with Histopaque-1077 (Sigma).
Circulating cells recruited by VEGF to the heart of 4- to 5-week-old
mice were isolated by digesting the perfused heart tissue with collage-
nase D (Roch 1088858) for 1 hr at 37ºC and centrifugation at 1200 rpm
for 7 min. For further fractionation of CD45+cells, CXCR4+cells, or
CD11b+cells, cells were reacted with the respective antibodies and puri-
fied (to 96% purity) through capture on anti-PE antibody-coated micro-
magnetic beads (Miltenyi Biotec). Antibodies used were as follows: anti-
mouse CD45 (PE-labeled, clone 30-F11 Pharmingen and APC-labeled,
clone 104 e-Bioscience), mAb against mouse CXCR4 (PE- or FITC-
labeled, clone 2B11 Pharmingen), mAb against mouse CD11b (PE-
labeled, clone M1/70 Pharmingen, M1/70 e-Bioscience), and isotype
Bone Marrow Transplantation Experiments
Double-transgenic mice (double-positive for Ly 5.1 and Ly 5.2) were ex-
posed to a single lethal dose of 10 Gy total body irradiation (TBI) from
a Gamma cell 40 by a Gamma beam Cesium 137 source (produced by
NDS Canada) at a 0.95 Gy/min dose rate. On the following day, the
mice were inoculated intravenously with 5 ? 106donor BM cells originat-
ing from B6.129S7-Gtrosa26 mice (single-positive for Ly 5.2) or from
C57BL/6-Tg (ACTbEGFP010sb mice). At 30 days posttransplant, the
level of donor-type chimerism was >89%, as evidenced by FACS analysis
of the donor single-positive marker (Ly 5.2) or of EGFP expression. This
was also confirmed by b-galactosidase staining or EGFP expression ob-
servation of the hematopoietic organs (bone marrow, spleen, and thy-
mus). VEGF expression was then switched on and the liver was retrieved
2 weeks thereafter.
Aortic Ring Sprouting Assay
Thoracic aortas were dissected from 8- to 10-week-old male mice, the
adventitia was removed, and 0.5 mm ‘‘rings’’ were embedded in collagen
as described by Licht et al. (2003). The collagen was then overlaid with
either a medium alone (BIO-MPM-1) or a medium containing the tested
factororcells ortheir conditioned medium,andthe plates were incubated
at 37ºC in a humidified 5% CO2atmosphere with triweekly medium re-
placement. The MMP inhibitor GM6001 (Chemicon CC1010) was added
at a concentrationof 25 mM. After 7 days, the rings were fixed in 4% form-
aldehyde for 24 hr, followed by staining with crystal violet (0.02%). Micro-
graphs of representative ringswere takenusing a digital camera and mor-
phometric analysis of sprouting was performed on four rings manually
using Image-J software according to Nissanov et al. (1995).
AMD3100 (a generous gift from AnorMED Inc.) was dissolved in 0.1M
NaHCO3(pH 7.4), and a total of 5 mg per mouse was continuously deliv-
ered with the aid of a subcutaneous miniosmotic pump (Alzet) over a pe-
riod of 14 days, starting 24 hr prior to switching on VEGF expression.
Control animals were similarly implanted with saline releasing pumps.
LAD Ligation and Myocardial Angiogenesis
Myocardial infarction was induced as previously described (Chimenti
et al., 2004).
Wound Healing Angiogenesis
NOD-SCID mice were anesthetized (Ketamine 50 mg/ml and xylazine
5 mg/ml, i.p.) and a 1 cm long, full thickness incision was made on the
dorsal skin. Incisions were closed by cyanoacrylate glue and examined
48 hr later. Half of the wound area was processed for RNA analysis; the
other half was fixed in 4% buffered paraformaldehyde and used for histo-
For other methods, see Supplemental Experimental Procedures.
Supplemental Data include five figures and Supplemental Experimental
Procedures and will be found with this article online at http://www.cell.
We thank Drs. Amnon Peled, Tamar Licht, Anat Globerman, and Dalit
May for their valuable contributions and the Israel Science Foundation
and the European Vascular Genomic Network (EVGN) for financial
Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc. 187
support. M.G. was supported by a postdoctoral fellowship from the Israel
Cancer Research Fund (ICRF) and from the Ministry of Science, Israel.
Received: September 12, 2004
Revised: August 11, 2005
Accepted: October 7, 2005
Published: January 12, 2006
Abbott, J.D., Huang, Y., Liu, D., Hickey, R., Krause, D.S., and Giordano,
F.J. (2004). Stromal cell-derived factor-1alpha plays a critical role in stem
cell recruitment to the heart after myocardial infarction but is not sufficient
to induce homing in the absence of injury. Circulation 110, 3300–3305.
Asahara, T., Murohara, T., Sullivan, A., Silver, M., van der Zee, R., Li, T.,
Witzenbichler, B., Schatteman, G., and Isner, J.M. (1997). Isolation of
putative progenitor endothelial cells for angiogenesis. Science 275,
Asahara, T., Masuda, H., Takahashi, T., Kalka, C., Pastore, C., Silver, M.,
Kearne, M., Magner, M., and Isner, J.M. (1999). Bone marrow origin of
endothelial progenitor cells responsible for postnatal vasculogenesis in
physiological and pathological neovascularization. Circ. Res. 85, 221–
Barleon, B., Sozzani, S., Zhou, D., Weich, H.A., Mantovani, A., and
Marme,D. (1996). Migration ofhuman monocytes inresponse tovascular
endothelial growth factor (VEGF) is mediated via the VEGF receptor flt-1.
Blood 87, 3336–3343.
Bergers, G., Brekken, R., McMahon, G., Vu, T.H., Itoh, T., Tamaki, K.,
Tanzawa, K., Thorpe, P., Itohara, S., Werb, Z., and Hanahan, D.
(2000). Matrix metalloproteinase-9 triggers the angiogenic switch during
carcinogenesis. Nat. Cell Biol. 2, 737–744.
Ceradini, D.J., Kulkarni, A.R., Callaghan, M.J., Tepper, O.M., Bastidas,
N., Kleinman, M.E., Capla, J.M., Galiano, R.D., Levine, J.P., and Gurtner,
G.C. (2004). Progenitor cell trafficking is regulated by hypoxic gradients
through HIF-1 induction of SDF-1. Nat. Med. 10, 858–864.
Chimenti, S., Carlo, E., Masson, S., Bai, A., and Latini, R.(2004).Myocar-
dial infarction: animal models. Methods Mol. Med. 98, 217–226.
De Clercq, E. (2003). The bicyclam AMD3100 story. Nat. Rev. Drug
Discov. 2, 581–587.
De Palma, M., Venneri, M.A., Roca, C., and Naldini, L. (2003). Targeting
exogenous genes to tumor angiogenesis by transplantation of genetically
modified hematopoietic stem cells. Nat. Med. 9, 789–795.
Dor, Y., Djonov, V., Abramovitch, R., Itin, A., Fishman, G.I., Carmeliet, P.,
Goelman, G., and Keshet, E. (2002). Conditional switching of VEGF pro-
vides new insights into adult neovascularization and pro-angiogenic ther-
apy. EMBO J. 21, 1939–1947.
Garcia-Barros, M., Paris, F., Cordon-Cardo, C., Lyden, D., Rafii, S.,
Haimovitz-Friedman, A., Fuks, Z., and Kolesnick, R. (2003). Tumor re-
sponse to radiotherapy regulated by endothelial cell apoptosis. Science
Geissmann, F., Jung, S., and Littman, D.R. (2003). Blood monocytes
consist of two principal subsets with distinct migratory properties. Immu-
nity 19, 71–82.
Hattori, K., Dias, S., Heissig, B., Hackett, N.R., Lyden, D., Tateno, M.,
Hicklin, D.J., Zhu, Z., Witte, L., Crystal, R.G., et al. (2001). Vascular endo-
thelial growth factor and angiopoietin-1 stimulate postnatal hematopoie-
sis by recruitment of vasculogenic and hematopoietic stem cells. J. Exp.
Med. 193, 1005–1014.
Infanger, M., Schmidt, O., Kossmehl, P., Grad, S., Ertel, W., and Grimm,
D. (2004). Vascular endothelial growth factor serum level is strongly en-
hanced after burn injury and correlated with local and general tissue
edema. Burns 30, 305–311.
Jung, S., Aliberti, J., Graemmel, P., Sunshine, M.J., Kreutzberg, G.W.,
Sher, A., and Littman, D.R. (2000). Analysis of fractalkine receptor
CX(3)CR1 function by targeted deletion and green fluorescent protein re-
porter gene insertion. Mol. Cell. Biol. 20, 4106–4114.
M., Li, T., Isner, J.M., and Asahara, T. (2000). Transplantation of ex vivo
expanded endothelial progenitor cells for therapeutic neovascularization.
Proc. Natl. Acad. Sci. USA 97, 3422–3427.
Kistner, A.,Gossen, M.,Zimmermann,F.,Jerecic, J.,Ullmer,C., Lubbert,
H., and Bujard, H. (1996). Doxycycline-mediated quantitative and tissue-
specific control of gene expression in transgenic mice. Proc. Natl. Acad.
Sci. USA 93, 10933–10938.
Kocher, A.A., Schuster, M.D., Szabolcs, M.J., Takuma, S., Burkhoff, D.,
Wang, J., Homma, S., Edwards, N.M., and Itescu, S. (2001). Neovascu-
larization of ischemic myocardium by human bone-marrow-derived an-
gioblasts prevents cardiomyocyte apoptosis, reduces remodeling and
improves cardiac function. Nat. Med. 7, 430–436.
Kollet, O., Shivtiel, S., Chen, Y.Q., Suriawinata, J., Thung, S.N., Dabeva,
M.D., Kahn,J., Spiegel,A., Dar, A., Samira, S., et al. (2003). HGF,SDF-1,
and MMP-9 are involved in stress-induced human CD34+ stem cell re-
cruitment to the liver. J. Clin. Invest. 112, 160–169.
Kraft, A., Weindel, K., Ochs, A., Marth, C., Zmija, J., Schumacher, P.,
Unger, C., Marme, D., and Gastl, G. (1999). Vascular endothelial growth
factor in the sera and effusions of patients with malignant and nonma-
lignant disease. Cancer 85, 178–187.
Licht, T., Tsirulnikov, L., Reuveni, H., Yarnitzky, T., and Ben-Sasson, S.A.
(2003). Induction of pro-angiogenic signaling by a synthetic peptide de-
rived from the second intracellular loop of S1P3 (EDG3). Blood 102,
Lyden, D., Hattori, K., Dias, S., Costa, C., Blaikie, P., Butros, L., Chad-
burn, A., Heissig, B., Marks, W., Witte, L., et al. (2001). Impaired recruit-
ment of bone-marrow-derived endothelial and hematopoietic precursor
cells blocks tumor angiogenesis and growth. Nat. Med. 7, 1194–1201.
Machein, M.R., Renninger, S., de Lima-Hahn, E., and Plate, K.H. (2003).
Minor contribution of bone marrow-derived endothelial progenitors to the
vascularization of murine gliomas. Brain Pathol. 13, 582–597.
Muller, A., Homey, B., Soto, H., Ge, N., Catron, D., Buchanan, M.E.,
McClanahan, T., Murphy, E., Yuan, W., Wagner, S.N., et al. (2001). In-
volvement of chemokine receptors in breast cancer metastasis. Nature
Nissanov, J., Tuman, R.W., Gruver, L.M., and Fortunato, J.M. (1995).
Automatic vessel segmentation and quantification of the rat aortic ring
assay of angiogenesis. Lab. Invest. 73, 734–739.
Ogawa, H., Suefuji, H., Soejima, H., Nishiyama, K., Misumi, K., Takazoe,
K., Miyamoto, S., Kajiwara, I., Sumida, H., Sakamoto, T., et al. (2000). In-
creased blood vascular endothelial growth factor levels in patients with
acute myocardial infarction. Cardiology 93, 93–99.
Orimo, A., Gupta, P.B., Sgroi, D.C., Arenzana-Seisdedos, F., Delaunay,
T., Naeem, R., Carey, V.J., Richardson, A.L., and Weinberg, R.A.
(2005). Stromal fibroblasts present in invasive human breast carcinomas
promote tumor growth and angiogenesis through elevated SDF-1/
CXCL12 secretion. Cell 121, 335–348.
Peled, A., Petit, I., Kollet, O., Magid, M., Ponomaryov, T., Byk, T., Nagler,
A., Ben-Hur, H., Many, A., Shultz, L., et al. (1999). Dependence of human
stem cell engraftment and repopulation of NOD/SCID mice on CXCR4.
Science 283, 845–848.
Peters, B.A., Diaz, L.A., Polyak, K., Meszler, L., Romans, K., Guinan,
E.C., Antin, J.H., Myerson, D., Hamilton, S.R., Vogelstein, B., et al.
(2005). Contribution of bone marrow-derived endothelial cells to human
tumor vasculature. Nat. Med. 11, 261–262.
Rafii, S., and Lyden, D. (2003). Therapeutic stem and progenitor cell
transplantation for organ vascularization and regeneration. Nat. Med. 9,
Rajantie, I., Ilmonen, M., Alminaite, A., Ozerdem, U., Alitalo, K., and
Salven, P. (2004). Adult bone marrow-derived cells recruited during
188 Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc.
angiogenesis comprise precursors for periendothelial vascular mural
cells. Blood 104, 2084–2086.
Rehman, J., Li, J., Orschell, C.M., and March, K.L. (2003). Peripheral
blood ‘‘endothelial progenitor cells’’ are derived from monocyte/macro-
phages and secrete angiogenic growth factors. Circulation 107, 1164–
S., Manova, K., Mittal, V., and Benezra, R. (2003). Effect of angiogenesis
inhibition by Id loss and the contribution of bone-marrow-derived endo-
thelial cells in spontaneous murine tumors. Cancer Cell 4, 277–289.
Salven, P., Manpaa, H., Orpana, A., Alitalo, K., and Joensuu, H. (1997).
Serum vascular endothelial growth factor is often elevated in dissemi-
nated cancer. Clin. Cancer Res. 3, 647–651.
Schachinger, V., Assmus, B., Britten, M.B., Honold, J., Lehmann, R.,
Teupe, C., Abolmaali, N.D., Vogl, T.J., Hofmann, W.K., Martin, H., et al.
inacute myocardial infarction: final one-year results of the TOPCARE-AMI
Trial. J. Am. Coll. Cardiol. 44, 1690–1699.
Sipkins, D.A., Wei, X., Wu, J.W., Runnels, J.M., Cote, D., Means, T.K.,
Luster, A.D., Scadden, D.T., and Lin, C.P. (2005). In vivo imaging of spe-
cialized bone marrow endothelial microdomains for tumour engraftment.
Nature 435, 969–973.
Takakura, N., Watanabe, T., Suenobu, S., Yamada, Y., Noda, T., Ito, Y.,
Satake, M., and Suda, T. (2000). A role for hematopoietic stem cells in
promoting angiogenesis. Cell 102, 199–209.
Wagers, A.J., Sherwood, R.I., Christensen, J.L., and Weissman, I.L.
(2002). Little evidence for developmental plasticity of adult hematopoietic
stem cells. Science 297, 2256–2259.
Yamaguchi, J., Kusano, K.F., Masuo, O., Kawamoto, A., Silver, M.,
Murasawa, S., Bosch-Marce, M., Masuda, H., Losordo, D.W., Isner,
J.M., and Asahara, T. (2003). Stromal cell-derived factor-1 effects on
ex vivo expanded endothelial progenitor cell recruitment for ischemic
neovascularization. Circulation 107, 1322–1328.
Yu, Z., Redfern, C.S., and Fishman, G.I. (1996). Conditional transgene
expression in the heart. Circ. Res. 79, 691–697.
Ziegelhoeffer, T., Fernandez, B., Kostin, S., Heil, M., Voswinckel, R.,
Helisch, A., and Schaper, W. (2004). Bone marrow-derived cells do
not incorporate into the adult growing vasculature. Circ. Res. 94,
Cell 124, 175–189, January 13, 2006 ª2006 Elsevier Inc. 189