Blockade of the Stromal Cell–Derived Factor-1/CXCR4 Axis Attenuates In vivo Tumor Growth by Inhibiting Angiogenesis in a Vascular Endothelial Growth Factor–Independent Manner

The chemokines are a group of chemoattractant cytokines that mediate several cellular functions. Stromal cell–derived factor-1α (SDF-1) is expressed by stromal cells, including fibroblasts and endothelial cells (1, 2), and interacts specifically with the seven-transmembrane, G protein–coupled receptor CXCR4 (3). CXCR4 is expressed in various cells, such as T lymphocytes, monocytes, neutrophils (4), and endothelial cells (5, 6). On ligand binding, CXCR4 activates several signaling cascades, including the phosphatidylinostiol 3-kinase and mitogen-activated protein kinase cascades, which induce cytoskeletal rearrangement, antiapoptosis effects, and cell growth (7). Importantly, the SDF-1/CXCR4 interaction is critical for the homing and retention of hematopoietic stem cells within the bone marrow and is essential in fetal hematopoiesis (8).

Recently, intensive research has indicated that CXCR4 is involved in increasing the metastatic potential of colon and breast cancer cells (2, 9, 10). For example, CXCR4 was one of the few genes that was up-regulated in bone-metastasized breast cancer cells (11), and cells that had metastasized to the lungs expressed very high levels of CXCR4 compared with the parental tumor cells. Another study found that SDF-1 was up-regulated in malignant tissues (12). In vivo, neutralizing the interaction of CXCR4 and SDF-1 significantly impaired the metastasis of breast cancer cells and cell migration (2). These findings suggest the possibility that inhibition of the SDF-1/CXCR4 axis could be a strategy for the prevention of cancer cell metastasis. The mechanism by which CXCR4 expression enhances tumor metastasis is still unclear; however, the activation of CXCR4 by SDF-1 seems to have the ability to trigger the adhesion of a variety of tumor cell lines to extracellular matrix substrates, such as fibronectin (13, 14), and to vascular endothelial cells by increasing the vascular permeability (12, 15).

Tumor neovascularization, a rate-limiting step in cancer progression, is thought to be established by the sprouting of blood vessels through the division of differentiated endothelial cells. However, the growth of new vessels can be mediated in several ways. Recently, circulating endothelial progenitor cells mobilized from the bone marrow were detected in the peripheral blood of several species and were implicated in the neoangiogenesis involved in tumorigenesis as well as in the formation of new vessels after trauma, burn injury, and myocardial infarction (16–18). The SDF-1/CXCR4 axis mediates the guidance of primordial stem cells to sites of rapid vascular expansion during embryonic organogenesis (19). An analysis of CXCR4-deficient mice revealed that the receptor was essential for fetal gastrointestinal vascular formation (20), suggesting a pivotal role of the SDF-1/CXCR4 axis in fetal angiogenesis (8). Like embryonic vasculogenesis, tumor angiogenesis might be mediated by various progenitor cells (21). We analyzed the contribution of the SDF-1/CXCR4 interaction to tumor neoangiogenesis using a mouse model. Here, we show that the in vivo neutralization of CXCR4 also results in the attenuation of tumor growth by the inhibition of tumor neovascularization in a vascular endothelial growth factor (VEGF)–independent manner regardless of whether the cancer cells express CXCR4. CXCR4-targeting therapy might be applied as an antiangiogenic strategy for treatment of a broad spectrum of cancers.

Animals and cells. All C57BL/6 mice and BALB/c nude mice were purchased from Clea Corp. (Tokyo, Japan). The transgenic mice (C57BL/6 background) that ubiquitously expressed enhanced green fluorescent protein (GFP mice) were a generous gift from Dr. Masaru Okabe (Osaka University, Osaka, Japan; ref. 22). All procedures involving experimental animals were done in accordance with protocols approved by the Institutional Committee for Animal Research of the University of Tokyo and complied with the USPHS Policy on Humane Care and Use of Laboratory Animals. Colon38 (23) and PancO2 (24) cells are colon and pancreatic cancer cells, respectively, derived from C57BL/6 mice. PancO2 cells were obtained from Dr. Michael A. Hollingsworth (Eppley Institute, University of Nebraska, Omaha, NE).

Stable knockdown of CXCR4. A plasmid carrying RNA interference targeted to mouse CXCR4 was constructed as described previously (25, 26). The siCXCR4 sequence of the mouse CXCR4 gene (5′-GCAAAGACTTATATAATATAT-3′) was selected using our original algorithm. Colon38 cells were transfected with pcPUR+U6-siCXCR4 or pcPUR+U6-siRenilla (control) and selected as puromycin-resistant pools. Then, quantitative reverse transcription-PCR (RT-PCR) was done to confirm the CXCR4 mRNA suppression using the primers 5′-TCAGCCTGGACCGGTACCT-3′ and 5′-GCAGTTTCCTTGGCCTTTGA-3′.

Bone marrow transplantation and tumor implantation model. The bone marrow of lethally irradiated C57BL/6 mice was reconstituted by transplantation with bone marrow cells from GFP mice (GFP-BMT mice). Briefly, wild-type C57BL/6 mice were lethally irradiated with a total dose of 950 rads (MBR-1520RB, Hitachi, Tokyo, Japan; ref. 27); then, 2 × 10⁶ bone marrow cells from GFP mice were injected into the tail veins of the irradiated recipient mice. The bone marrow cells of the GFP-BMT mice were sampled at 4 weeks after bone marrow transplantation, and the degree of chimerism was investigated by flow cytometry (EPICS XL, Beckman Coulter, Fullerton, CA). More than 85% of the cells in the recipient bone marrow were GFP positive using these experimental conditions (data not shown). Tumors were induced by s.c. injection of 2 × 10⁶ cancer cells into the flank >4 weeks after the bone marrow transplantation. Large tumors were typically observed by 4 weeks after tumor implantation. The mice were anesthetized with pentobarbital sodium (120 mg/kg), and the tumor tissues were harvested for histologic analysis.

Isolation of tumor-infiltrating cells and reverse transcription-PCR analysis. Tumor-infiltrating cells (TIC) were isolated from the tumors formed by Colon38 and PancO2 cells using density gradient centrifugation in Percoll/RediGrad (Amersham Biosciences, Buckinghamshire, United Kingdom) as described previously (28). The total RNA was extracted from the TICs using ISOGEN reagent (Nippon Gene Co., Tokyo, Japan), and the reverse transcription reaction and PCR amplification were done as described (29). The PCR primer sequences were as follows: sense 5′-GGCTGTAGAGCGAGTGTTGC-3′ and antisense 5′-GTAGAGGTTGACAGTGTAGAT-3′ for CXCR4 (29) and sense 5′-GTTGGATACAGGCCAGACTTTGTTG-3′ and antisense 5′-GATTCAACTTGCGCTCATCTTAGGC-3′ for hypoxanthine phosphoribosyltransferase (HPRT).

Immunohistochemistry. Tumor tissue samples were fixed in 2% paraformaldehyde and embedded with a Technovit catalyst system (Heraeus Kulzer GmbH & Co. KG, Wehrheim, Germany; ref. 30). The primary antibodies were as follows: rabbit anti-rat CXCR4 (Torrey Pines Biolabs, Inc., Houston, TX), rabbit anti-mouse CXCR4 (H-118) and anti-VEGF (sc-507; Santa Cruz Biotechnology, Santa Cruz, CA), rabbit anti-mouse SDF-1α (Torrey Pines Biolabs), rat anti-mouse CD31 (BD PharMingen, San Diego, CA), and rat anti-mouse Mac3 (BD PharMingen). The secondary antibodies were conjugated with fluorochrome Alexa Fluor 488 or 555 (Molecular Probes, Inc., Eugene, OR). The sections were observed under a confocal microscope (Leica Microsystems, Wetzlar, Germany; ref. 31).

In vivo neutralizing antibody studies. Cancer cells (Colon38 or PancO2; 8 × 10⁶) were s.c. injected into BALB/c nude mice. One group of tumor-bearing mice (n = 5) then received an i.p. injection of 10 μg rabbit anti-rat CXCR4-neutralizing antibody, which is reported to also bind to murine CXCR4 (32). The control group of tumor-bearing mice received 10 μg normal rabbit IgG. The mice were treated every 24 hours starting on day 3 for a total of eight separate injections of anti-CXCR4 antibody. The tumor size was measured, and the volume was calculated as [length (mm) × width (mm)²] / 2. The experiments were also done using C57BL/6 mice (n = 4).

3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay. The direct effect of the neutralizing antibody on the viability of the tumor cells was assessed by seeding 3 × 10⁴ Colon38 cells in 24-well microplates and replacing the medium with medium containing 10 μg/mL normal IgG or 10 μg/mL anti-CXCR4-neutralizing antibody (n = 3) after 24 hours. The number of viable cells was determined using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay (Sigma, St. Louis, MO) at 24, 48, and 72 hours (33). Similarly, a total of 3 × 10⁴ Colon38-siCXCR4 or Colon38-siRenilla cells were cultured with 300 ng/mL recombinant human SDF-1α (PeproTech EC, London, United Kingdom) in 24-well microplates, and the number of viable cells was analyzed using the MTT assay at 24, 48, and 72 hours.

Fluorescent phalloidin staining. A total of 10⁵ Colon38 or PancO2 cells were seeded in two-well chambers and incubated with 10 μg/mL normal IgG or 10 μg/mL anti-CXCR4-neutralizing antibody with 300 ng/mL recombinant human SDF-1α for 24 hours; then, the cells were stained with Alexa Fluor 488–labeled phalloidin according to the manufacturer's instructions.

Cell migration assay (wound closure assay). A total of 10⁵ Colon38 or PancO2 cells were seeded in two-well chambers, and confluent cell monolayers were wounded by scraping using a pipette tip of the same width and replacing the medium with medium containing 10 μg/mL normal IgG or 10 μg/mL anti-CXCR4-neutralizing antibody with 300 ng/mL recombinant human SDF-1α for 24 hours. Then, the cells were fixed and stained with Diff-Quick, and cell migration was observed using bright-field microscopy at ×40 magnification.

Quantitative reverse transcription-PCR. A total of 5 × 10⁵ Colon38 or PancO2 cells were seeded in 6-cm dishes and incubated with 10 μg/mL normal IgG or 10 μg/mL anti-CXCR4-neutralizing antibody for 24 hours. The total RNA was extracted, treated with DNase, and purified. Quantitative RT-PCR analysis was done using an ABI 7000 Real-time PCR System (Applied Biosystems, Foster City, CA). The mRNA level of each gene was normalized to HPRT. The primers were as follows: basic fibroblast growth factor (bFGF), forward 5′-CACCAGGCCACTTCAAGGA-3′ and reverse 5′-GATGGATGCGCAGGAAGAA-3′; platelet-derived growth factor (PDGF), forward 5′-AAGCTCGGGTGACCATTCG-3′ and reverse 5′-ACTTTCGGTGCTTGCCTTTG-3′; and placenta growth factor (PlGF), forward 5′-CCCTGTCTGCTGGGAACAA-3′ and reverse 5′-GCTGCGACCCCACACTTC-3′. The SDF-1, matrix metalloproteinase-9 (MMP-9), intercellular adhesion molecule (ICAM), and VCAM (VCAM) primer sequences were reported previously (34).

Late-outgrowth endothelial colony assay. To isolate the peripheral blood mononuclear cells (PBMC; ref. 18), blood samples (500-1,000 μL) from mice were collected in heparinized tubes ∼3 weeks after tumor implantation. The PBMCs were isolated by Ficoll gradient centrifugation (Amersham Biosciences AB, Uppsala, Sweden). To detect circulating endothelial cells, 3 × 10⁵ freshly isolated PBMCs were cultured in modified endothelial growth medium (EGM), which was composed of X-vivo-20 serum-free medium with VEGF (10 ng/mL, R&D, Minneapolis, MN), endothelial cell growth supplement (30 μg/mL, Upstate, Lake Placid, NY), human recombinant bFGF (5 ng/mL, Invitrogen, Carlsbad, CA), heparin (5 units/mL), streptomycin (100 μg/mL), penicillin (100 units/mL), and fungizone (0.25 μg/mL). The PBMCs were placed in two-well chambers coated with 0.2% gelatin (17, 18) and incubated at 37°C in a humidified environment with a 5% CO₂ atmosphere. Monocytes and mature endothelial colonies attached to the well chambers within 3 days. The nonadherent cells were transferred to other wells in EGM after 3 days. After 2 weeks, the endothelial colonies were characterized by the metabolic uptake of acetylated low-density lipoprotein labeled with 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI-Ac-LDL; Biomedical Technologies, Stoughton, MA). The cells were incubated with 10 μg/mL DiI-Ac-LDL at 37°C for 4 hours and examined by fluorescence microscopy. DiI-Ac-LDL and GFP double-positive colonies that formed at 2 weeks were considered late-outgrowth colonies (17).

Measurement of local blood perfusion in superficial tumor tissue. The blood perfusion of the tumors was measured using a laser Doppler perfusion imaging (LDPI) system (Moor Instruments, Devon, United Kingdom; refs. 35, 36). The LDPI data were acquired from mice on day 10 following five antibody treatments given at 24-hour intervals. Each mouse was anesthetized 15 minutes before the recordings to eliminate artifacts caused by body movements; the mouse was placed on a heating plate at 40°C, and the LDPI recording was made (35, 36).

Estimation of vascular endothelial growth factor levels in subcutaneuse tumors and cultured supernatants. The tumor tissues were harvested on day 14 after five antibody treatments and homogenized with a tissue homogenizer in 9 volumes of lysis buffer [300 mmol/L NaCl, 15 mmol/L Tris-HCl, 2 mmol/L MgCl₂, 2 mmol/L Triton X-100, 20 ng/mL pepstatin A, 20 ng/mL leupeptin, 20 ng/mL aprotinin (pH 7.4)]. A total of 3 × 10⁵ Colon38 or PancO2 cells were seeded in six-well plates and incubated with fetal bovine serum–free medium containing 10 μg/mL normal IgG or 10 μg/mL anti-CXCR4-neutralizing antibody for 24 hours. The ELISA assay for VEGF was done by the SRL analysis service (Tokyo, Japan).

Statistical analysis. The data are expressed as mean ± SE. Comparisons between the groups were analyzed using Student's t test. P < 0.05 was considered statistically significant.

CXCR4-positive cells contribute to the establishment of tumor tissues in a cancer cell type–independent manner. To estimate the role of the SDF-1/CXCR4 axis in the establishment of gastrointestinal tumors, in addition to its known metastasis-promoting ability, we analyzed two models of s.c. tumors in mice using two mouse-derived cancer cells, Colon38 and PancO2. CXCR4 mRNA was not detected in cultured PancO2 cells but was detected in Colon38 cells (Fig. 1A). However, CXCR4-positive cells were detected in the tissues of tumors established from implanted Colon38 and PancO2 cells; the CXCR4-positive cells were observed around the tumor vessels and occasionally in the endothelium (Fig. 1B,, arrow). These findings suggested that the cells infiltrating the tumor might express CXCR4 regardless of whether the cancerous cells themselves express CXCR4. To clearly distinguish cancer cells from TICs, we established a mouse model in which the bone marrow was depleted by irradiation and then reconstituted by transplantation of GFP-tagged bone marrow cells (GFP-BMT mice). Many bone marrow–derived cells were found to infiltrate into both types of tumor tissues, and GFP and CXCR4 double-positive cells were detected around the tumor vessels (Fig. 1C,, arrow). To confirm the expression of CXCR4 in the TICs, TICs were isolated from tumor tissues as described previously, and the expression levels of various chemokine receptors were analyzed by RT-PCR (data not shown). CXCR4 mRNA was detected in the TICs of both Colon38 and PancO2 tumors (Fig. 1A). To investigate the percentage of CXCR4-expressing cells in bone marrow–derived cells, we counted the total numbers of GFP and CXCR4 double-positive cells as the number per square millimeter using fluorescence microscopy (n = 3). GFP and CXCR4 double-positive cells constituted ∼3.4% (32.3 ± 2.1 versus 961.3 ± 36.8) or 2.1% (65.7 ± 4.6 versus 3185.7 ± 149.8) of all bone marrow–derived cells in Colon38-derived or PancO2-derived tumors, respectively. For a while, SDF-1 was detected in the extracellular portion of the stromal area, especially around the vessels in both Colon38-derived and PancO2-derived tumors (data not shown). To quantify the expression of SDF-1, we extracted total RNA from s.c. tissues and the two cell lines and did quantitative RT-PCR (Fig. 1D). The expression of SDF-1 was induced in both tumor tissues. PancO2 cells did not express SDF-1 in vitro; nevertheless, the level of SDF-1 expression in s.c. tumors was not very different from that in Colon38. These findings suggest that SDF-1 is secreted from noncancerous tissues in the tumors or expression could be up-regulated in the tumor tissues. Therefore, the SDF-1/CXCR4 axis seems to contribute to the establishment of tumor tissues via the expression of CXCR4 on infiltrating cells regardless of whether the cancer cells themselves express CXCR4.

CXCR4 neutralization prevents the growth of Colon38 and PancO2 tumors regardless of CXCR4 expression by the cancer cells. We evaluated the therapeutic potential of the neutralization of CXCR4 for the inhibition of tumor formation using the tumor transplant model. To interfere with SDF-1/CXCR4 signaling, BALB/c nude mice transplanted with Colon38 and PancO2 cells were treated with anti-CXCR4-neutralizing antibody or a control antibody using the dose schedule described in a previous report (32). The growth of Colon38 xenograft tumors was clearly suppressed in the group treated with the neutralizing antibody compared with the control group (n = 5; Fig. 2A). Neutralizing antibody against CXCR4 also suppressed the growth of PancO2 tumors (n = 5; Fig. 2B). This finding was reproduced in the experiment using C57BL/6 mice. As shown in Fig. 2C, neutralizing antibody against CXCR4 also suppressed the growth of Colon38 tumors in C57BL/6 mice (n = 4). As Colon38 cells had been shown to express CXCR4 (Fig. 1A), we examined whether the anti-CXCR4-neutralizing antibody could directly inhibit their growth. Colon38 cells were cultured in the presence of 10 μg/mL anti-CXCR4 antibody or control antibody to simulate the concentration in the peripheral blood of treated mice. Under these in vitro conditions, anti-CXCR4 antibody treatment had no effect on the growth of Colon38 cells (n = 3; P = 0.93; Fig. 2D). To confirm that the effect was independent of the CXCR4 expression by cancer cells themselves, we established Colon38-siCXCR4 cells in which the CXCR4 gene was stably suppressed. As shown in Fig. 2E, the siCXCR4 effectively blocked CXCR4 mRNA expression. Colon38-siCXCR4 cells or Colon38-siRenilla cells were transplanted s.c. into mice and the difference in growth rates was compared. As shown in Fig. 2G, growth rates were not significantly different between the two groups (n = 5; P > 0.1). Growth rates were also similar in the in vitro culture experiments (Fig. 2F).

In addition, we investigated whether the neutralizing antibody has other biological effects in vitro, because the SDF-1/CXCR4 axis is significant in breast and colon cancer metastasis. Obvious cytoskeletal changes were not detected by fluorescent phalloidin staining of groups of the two cell lines stained with anti-CXCR4 antibody and control antibody (data not shown). To examine whether the anti-CXCR4-neutralizing antibody treatment affects cell migration, we did a cell migration assay (wound closure assay). No difference in cell migration was observed between the groups for either cell line (data not shown). Between the two groups for both cell lines, there was no statistical difference in the mRNA expression of ICAM (n = 4; P > 0.1), VCAM (n = 4; P > 0.1), PDGF (n = 4; P > 0.1), PlGF (n = 4; P > 0.1), bFGF (n = 4; P > 0.1), or MMP-9 (n = 4; P > 0.1). The VEGF secretion of the Panco2 cells in vitro was not statistically different between the two groups (n = 4; P = 0.92). VEGF production was not detected in the Colon38 cell line. These findings indicate that the suppression of tumor growth was not mediated by the direct inhibition of cancer cell growth.

CXCR4 neutralization decreases the development of tumor endothelium in vivo. As CXCR4-positive cells were detected among endothelial cells as well as in the perivascular area (Fig. 1B), the effect of CXCR4 neutralization on tumor angiogenesis was estimated by histologic examination of tumors for CD31, an endothelial marker (Fig. 3A). The capillary density was calculated as the number of capillaries per square millimeter exhibiting expression of CD31 based on counts in 10 randomly selected fields from each tissue preparation examined by confocal microscopy. Staining with anti-CD31 antibody showed that the density of vessels in the tumors of the mice treated with the neutralizing antibody was significantly lower than that in the tumors of the control group (n = 3; P = 0.00048; Fig. 3B). In addition, the Colon38-siCXCR4 or Colon38-siRenilla cells were transplanted into mice s.c. and the vascularity of the tumors was compared. The capillary densities in the tumors in the two groups were not significantly different (n = 3; P = 0.51; Fig. 3B). Recently, bone marrow–derived endothelial cells were reported to be involved in tumor angiogenesis in tumor implantation models (16–18, 37). To investigate whether bone marrow–derived endothelial cells formed the tumor vessel endothelium in our model, the expression of GFP in the endothelium of the tumor xenografts in GFP-BMT mice was analyzed. GFP and CD31 double-positive cells were observed lining the vessels only in rare instances, suggesting that the population of bone marrow–derived tumor endothelial cells was very limited in our model (Fig. 4A). The presence of endothelial progenitor cells in the peripheral blood of tumor-bearing mice was investigated using the late-outgrowth assay. PBMCs were cultured to permit the growth of endothelial colonies, which were then identified by the metabolic uptake of DiI-Ac-LDL after 2 weeks of culture. The expression of GFP was detected in most of the DiI-Ac-LDL-positive colonies (Fig. 4B), but these cells did not form large colonies under our experimental conditions. Importantly, late-outgrowth cells identified in vitro expressed the chemokine receptor CXCR4 (Fig. 4C).

In vivo neutralization of CXCR4 decreases the intratumor blood flow. As the blockade of CXCR4 induced a decrease in the development of intratumor endothelial cells, we monitored dynamic changes in the blood perfusion of the tumors using a LDPI system (35, 36). LDPI can provide noninvasive analysis of local perfusion in superficial tissues. The technique is based on the principle that the wavelength of laser light changes (Doppler shift) when it is reflected from a moving object (RBC in this case), whereas the wavelength of light reflected from a stationary object remains unchanged. The Doppler-shifted light is converted into an arbitrary perfusion signal, which is approximately proportional to the mean blood cell velocity multiplied by the concentration of moving blood cells (Fig. 5A). The blood flow in the tumors treated with neutralizing antibody was decreased to ∼65% of that in the control tumors for both s.c. Colon38 and PancO2 tumors (n = 7; P < 0.01 and P < 0.001, respectively; Fig. 5B). Taken together with the observations presented in Fig. 2, these findings indicate that neutralization of CXCR4 suppresses tumor growth by an antiangiogenic mechanism and not by direct inhibition of cell growth. In addition, the Colon38-siCXCR4 or Colon38-siRenilla cells were transplanted into mice s.c. The Doppler flow rates in the tumors were similar (n = 5; P = 0.57; Fig. 5B). These results indicate that the decrease in tumor vascularity with anti-CXCR4 neutralizing antibodies is not caused by a direct effect against cancer cells.

CXCR4 neutralization did not change vascular endothelial growth factor expression in the tumor tissues or induce critical anemia. Cells of inflammatory cell lineages that infiltrate into tumors have been reported to secrete VEGF, a pivotal angiogenic factor (38, 39). By immunohistochemistry, some of the infiltrated cells were Mac3 positive, which indicated that monocytes/macrophages had been recruited into the tumor tissues from the bone marrow (Fig. 6A). In a short time, few anti–smooth muscle actin–positive cells were detected in our analysis (data not shown), which might be consistent with reports that tumor microvessels often lack a lining of smooth muscle cells unlike normal vessels. To determine whether the TICs secrete VEGF in our xenograft tumor model, we counterstained the GFP-positive cells of Colon38 tumor with anti-VEGF antibody. As shown in Fig. 6B, VEGF was expressed in the GFP-positive, bone marrow–derived infiltrating cells in our xenograft tumors (Fig. 6B). To clarify the effect of blocking CXCR4 on VEGF expression, the VEGF concentrations in the tumors were determined using ELISA and compared between the groups treated with CXCR4-neutralizing antibody and control antibodies. As shown in Fig. 6C, the VEGF concentration was not significantly affected by CXCR4 neutralization. These findings suggest that the suppression of angiogenesis by the neutralization of CXCR4 is not attributable to a change in the VEGF concentration and that VEGF secretion does not always depend on CXCR4-positive TICs. Indeed, the blockade of CXCR4 did not cause a major change in the number of perivascular bone marrow–derived GFP-positive cells in the tumor tissues (Fig. 6D). Inflammatory cells that infiltrate tumors have been reported to express various angiogenic factors. It has been suggested that the cells producing vascular mitogens can be mobilized by other mechanisms independent of SDF-1/CXCR4 signaling. Furthermore, because the SDF-1/CXCR4 axis is indispensable for hematopoietic stem cell homing (8), the inhibition of this interaction might impair bone marrow hematopoiesis. To exclude the possibility that the decrease in tumor perfusion measured in mice treated with CXCR4-neutralizing antibody was caused by anemia, which can result in the underestimation of intratumor blood flow by the LDPI system, the peripheral blood cell counts were determined for each group (Table 1). The peripheral blood cell counts were not significantly different between the groups, which indicated that CXCR4 antibody did not induce a critical suppression of the bone marrow under our experimental conditions.

In this study, we showed that the SDF-1/CXCR4 axis, which is indispensable for hematopoiesis and angiogenesis in the embryo (20, 40), plays a pivotal role in tumor progression through promoting tumor neovascularization. Our findings might provide new insight on the significance of the SDF-1/CXCR4 axis in local tumor progression. The reported ability of CXCR4 neutralization to block the growth of metastatic lesions might be mediated in part by this inhibition of angiogenesis. Although this antigrowth effect is independent of CXCR4 expression by cancer cells, based on the results of small interfering RNA experiments, this does not contradict the recent reports that in vivo breast cancer growth was dependent on CXCR4 (41). The significance of the SDF-1/CXCR4 axis for cancer cell growth in vivo could differ by cancer cell type. For example, the CXCR4 expression of gastrointestinal tumors might be less than that of breast cancer. In addition, another mitogenic signaling pathway could compensate for the growth disadvantage by inhibiting the SDF-1/CXCR4 axis in gastrointestinal tumors. Indeed, oncogenic mutations of the K-Ras or B-Raf genes that strongly induce hyperproliferative capacity are often reported in gastrointestinal tumors. Importantly, the anticancer potential of the CXCR4-blocking strategy may be effective for a broad spectrum of cancers.

During tumor progression, infiltrating cells produce several potent angiogenic growth factors, cytokines, and proteases (38, 39). The recruitment and infiltration of circulating cells are mediated by members of the chemokine family of chemoattractive cytokines. In our murine tumor models, the neutralization of CXCR4 did not change the concentration of VEGF in the tumors, suggesting that other chemokine systems function in the recruitment of VEGF-secreting cells. Although the capillary density was lower in anti-CXCR4-treated tumors in spite of unchanged VEGF concentrations, our data do not exclude the significance of VEGF in tumor angiogenesis. Our results indicate that the SDF-1/CXCR4 axis does not always regulate tumor angiogenesis in a VEGF-dependent manner; for example, the SDF-1/CXCR4 axis might contribute to functional vascular establishment by the regulation of endothelial tube formation (1).

Recently, circulating endothelial progenitor cells mobilized from bone marrow have been detected in the peripheral blood of several species and shown to be involved in neoangiogenesis in tumors as well as in the formation of new vessels after trauma, burn injury, and myocardial infarction (16–18). We have already documented the roles of bone marrow–derived vascular progenitor cells in vascular remodeling using the original reconstituted bone marrow mouse model (27, 30). In the present study using Colon38 and Panco2 cells, bone marrow–derived endothelial cells were infrequently detected in the capillaries of the tumors. However, our findings do not exclude the possibility that SDF-1/CXCR4 axis neutralization inhibits bone marrow–dependent tumor vasculogenesis, as is the case for embryonic vasculogenesis (42), because peripheral blood–derived late-outgrowth colonies expressed CXCR4 (Fig. 4C; ref. 43). The proportion of bone marrow–derived endothelial cells incorporated during neovascularization might differ among tumor types (44, 45) or might be influenced by the local expression profiles of various cytokines or growth factors (46). A recent study reported that in vivo expression of SDF-1 in ischemic tissues and CXCR4-positive progenitor recruitment were enhanced by the transcription factor hypoxia-inducible factor-1 (HIF-1; ref. 47). In another report, the metastatic ability of cancer cells was regulated by HIF-1-dependent CXCR4 expression (48). These findings suggest that HIF-1 expression in tumors might affect the recruitment of CXCR4-positive endothelial progenitors. Therefore, the degree of angiogenic inhibition by CXCR4 neutralization might increase in proportion to the contribution made by the recruitment of bone marrow–derived CXCR4-positive progenitors. Further experiments should be conducted in tumors of different origins to analyze the variation in the contribution of bone marrow–dependent vasculogenesis to tumor angiogenesis and the antitumor effects of the blockade of CXCR4.

Inhibitors of angiogenesis, such as anti-VEGF antibody, are expected to be able to suppress the advancement of tumor growth (49, 50). Treatment with bevacizumab, a monoclonal antibody against VEGF, in combination with fluorouracil/leucovorin treatment resulted in higher response rates and longer median survival times than treatment with fluorouracil/leucovorin alone (51). Moreover, because neovascularization processes are not continuously active in adult tissues, the targeting of vasculogenic reactions would be a relatively tumor-specific therapy (52). In contrast to the strategies employed by many anticancer drugs, strategies that target specific molecules might induce selective effects against cancer tissues. For example, the actions and interactions of endothelial cells and pericytes in tumors are qualitatively different from those in normal tissues (53), which might permit the specific targeting of the tumor vasculature. Indeed, SU6668, an inhibitor of VEGF and PDGF receptors, disrupted the association of pericytes with endothelium and reduced the vascularity in tumor tissues only (54).

Our experiments showed that injection of anti-CXCR4 antibody caused no critical bone marrow suppression or ischemic event. Based on our data, the inhibition of vasculogenesis by CXCR4 neutralization might be cooperatively effective against tumors in combination with other angiogenic inhibitors, such as anti-VEGF antibody or chemotherapeutic drugs. In the future, combined strategies that include targeting SDF-1/CXCR4 might be promising anticancer therapies against a broad spectrum of cancers.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Mitsuko Tsubouchi and members of the Sata Laboratory for their technical assistance.