Inhibition of tumor-induced neovascularization appears to be an effective anticancer approach, although long-term angiogenesis inhibition may be required. An alternative to chronic drug administration is a gene therapy-mediated approach in which long-term in vivo protein expression is established. We have tested this approach by modifying murine bone marrow-derived cells with a gene encoding an angiogenesis inhibitor: a soluble, truncated form of the vascular endothelial growth factor receptor-2, fetal liver kinase-1 (Flk-1). Murine bone marrow cells were transduced with a retroviral vector encoding either truncated, soluble Flk-1 (tsFlk-1) together with green fluorescent protein (GFP) or GFP alone. Tumor growth in mice challenged 3 months after transplantation with tsFlk-1-expressing bone marrow cells was significantly inhibited when compared with tumor growth in control-transplanted mice. Immunohistochemical analysis of tumors in each group demonstrated colocalization of GFP expression in cells staining with endothelial cell markers, suggesting that the endothelial cells of the tumor-induced neovasculature were derived, at least in part, from bone marrow precursors. These results suggest that long-term expression of a functional angiogenesis inhibitor can be generated through gene-modified, bone marrow-derived stem cells, and that this approach can have significant anticancer efficacy. Modifying these cells seems to have the added potential benefit of targeting transgene expression to the tumor neovasculature, because bone marrow-derived endothelial cell precursors seem to be recruited in the process of tumor-induced angiogenesis.

Angiogenesis, the formation of new capillary blood vessels, is required for the growth and spread of cancer (1, 2, 3). The corollary to this is that inhibition of tumor-induced neovascularization may be an effective anticancer approach that could be broadly applicable because most, if not all, malignancies seem to be angiogenesis-dependent (4, 5, 6). Although an increasing number of studies have demonstrated the efficacy of recombinant antiangiogenic proteins in murine tumor models, a gene therapy-mediated approach to their delivery has a number of potential advantages (7, 8, 9). (a) Angiogenesis inhibition is likely to be a cytostatic therapy; therefore, long-term delivery of these agents, as can be achieved through a gene therapy-mediated approach, may be necessary. (b) Difficulties with protein production and maintenance of function, especially when “scaling up” for clinical trials, may be avoided by in situ expression in host tissues. And (c), continuous low levels of these proteins, as would be generated from gene-modified cells, may be the optimal delivery schedule (10).

Gene therapy strategies for tumor antiangiogenesis have, in fact, already been tested in a number of different murine tumor models, and this approach has had some success. Most of these studies have used either retroviral vector producers, naked DNA, or adenoviral vectors as the angiogenesis inhibitory gene-delivery vehicles (9). Unfortunately, retroviral vector producers may be impractical for human use, and only transient expression of the various proteins was established by the latter two approaches in which, after an initial delay, tumor growth often resumed. This probably occurred because antiangiogenic therapy is only cytostatic; tumor growth may resume once the restrictions of angiogenesis inhibition are removed. Therefore, long-term expression of angiogenesis inhibitors is likely to be required for sustained anticancer efficacy. Alternative gene therapy approaches are needed to ensure long-term expression of these proteins. In addition, however, the effects of chronic angiogenesis inhibition on physiological, angiogenesis-dependent processes, such as healing and fertility, will also have to be critically evaluated.

The use of bone marrow-derived cells as targets for transduction and, therefore, the source of angiogenesis inhibitor expression is intriguing for several reasons. (a) Transduction of bone marrow-derived stem cells with retroviral vectors has been shown to be feasible and results in long-term transgene expression from mature bone marrow-derived cells in mice (11). (b) Because bone marrow transplantation is a standard part of treatment for many high-risk malignancies in children, this approach could, ultimately, be readily incorporated into existing pediatric treatment protocols. And (c), the endothelial cells incorporated into tumor-induced new blood vessels may, in fact, be derived from precursor cells of bone marrow origin (12), and modifying these cells to express an inhibitor of endothelial cell activation may prevent endothelial cell precursor differentiation and/or create a milieu of angiogenesis inhibition at the sites where tumors are trying to induce neovascularization.

VEGF3 is one of the primary tumor-expressed endothelial cell mitogens (13). The activity of this ligand can be inhibited by a soluble, truncated form of one of its receptors, Flk-1/KDR (VEGFR-2; 14, 15, 16). We have shown previously that autocrine expression of truncated, soluble Flk-1 (tsFlk-1) from either ex vivo or in situ gene-modified neuroblastoma cells results in a significant inhibition of local angiogenesis and tumor growth in vivo(17, 18). The purpose of this study was to determine whether gene-modified bone marrow cells expressing this angiogenesis inhibitor could, through paracrine delivery of this angiogenesis inhibitor, provide the same antitumor efficacy.

Cell Lines.

The murine neuroblastoma cell line NXS2 (19), provided by Dr. R. Reisfeld (La Jolla, CA), was maintained in DMEM (Mediatech, Inc., Herndon, VA) supplemented with 10% fetal bovine serum (Summit Biotechnology, Ft. Worth, CO), 100 units/ml penicillin-100 μg/ml streptomycin (Life Technologies, Inc., Grand Island, NY), and 2 mml-glutamine (Life Technologies, Inc). GP+E86 (3T3-based retroviral packaging cell line; ATCC, Manassas, VA), SK-NEP-1 (human Wilms’ tumor cell line; American Type Culture Collection), and 293T cells (human embryonic kidney cells expressing SV40 large T antigen; Ref. 20) were maintained in medium similar to that of NXS2. HUVECs were obtained from Clonetics (Walkersville, MD) and maintained in Endothelial Growth Medium.

Making of Retroviral Vector Plasmids and Producer Cells.

Retroviral vector plasmids based on the murine stem cell virus and encoding either tsFlk-1 linked via an internal ribosomal entry site to GFP or GFP alone were constructed as described previously (17). Briefly, cDNA for the truncated, soluble VEGF receptor was provided by Dr. Pengnian Lin (Durham, NC). The tsFlk-1 cDNA was contained within an AdExFlk.6His plasmid that has been described previously (16). A 2.3-kb HindIII-BamHI fragment containing the tsFlk-1 transgene was excised from this vector and ligated into pSP72 (Promega, Madison, WI), cut with HindIII and BamHI to make p72-mFlk. A 2352-bp EcoRI-XhoI fragment from p72-mFlk was then ligated into MSCV-I-GFP cut with EcoRI and XhoI. MSCV-I-GFP is a retroviral expression plasmid that contains the MSCV long terminal repeat driving expression of enhanced GFP (CLONTECH, Palo Alto, CA). The resulting retroviral expression plasmid, MSCV-tsFlk-1-I-GFP, contains the GFP gene linked to an internal ribosomal entry site from the encephalomyocarditis virus 3′ of the tsFlk-1 cDNA. The MSCV-long terminal repeat that acts as a promoter in this construct is not an endothelial cell promoter and, therefore, both tsFlk-1 and GFP will be expressed in all transduced cells.

GP+E86 retroviral packaging cells were then transduced with conditioned medium containing high-titer, vesicular stomatitis virus-G-pseudotyped MSCV-tsFlk-1-I-GFP or MSCV-I-GFP vector particles supplemented with 6 μg/ml Polybrene (Sigma Chemical Co. Chemical, St. Louis, MO). These conditioned media were derived by cotransfection of 293T cells with the respective retroviral vector plasmids and two helper plasmids, one containing the gag and pol retroviral genes and the other containing the gene for the vesicular stomatitis virus-G envelope, as described previously (17). Transduced GP+E86 cells were then sorted by FACS (Becton Dickinson, Bedford, MA) to delete untransduced cells. Once expanded, viral production from these cells was determined by titering conditioned medium on NIH 3T3 cells (American Type Culture Collection).

Bone Marrow Cell Transduction and Transplantation.

Retroviral transduction of murine bone marrow cells was performed as described previously (11). Briefly, bone marrow was harvested by flushing the femurs and tibias of either 8- to 12-week-old female A/J mice (Jackson Laboratory, Bar Harbor ME) or C.B-17 SCID mice (Charles River Laboratory, Wilmington, MA) with 2% heat-inactivated FCS (Hyclone, Logan, UT) in PBS,2 days after i.p. injection of 150 mg/kg 5-fluororacil (Pharmacia, Kalamazoo, MI). The marrow cores were dissociated by flushing through a 21-gauge needle, and the cells were counted after red-cell lysis. Marrow cells were then stimulated for 48 h with 20 ng/ml mouse IL-3, 50 ng/ml human IL-6, and 50 ng/ml mouse stem cell factor (R & D Systems, Minneapolis, MN) in DMEM supplemented with 15% heat-inactivated FCS. Bone marrow cells were subsequently cocultured with irradiated (1200 cGy; Cesium-137 source; Gammacell 40 Exactor; MDS Nordion, Inc., Ontario, Canada) viral producer cells using the above culture medium supplemented with 6 μg/ml Polybrene. This coculture was performed on gelatin-coated plates to prevent detachment of the irradiated producer cells. Forty-eight h later, nonadherent bone marrow cells were gently rinsed off the viral producer-cell monolayers, pelleted, and resuspended in PBS for transplantation. Two million gene-modified bone marrow cells were then administered via tail vein to recipient female A/J or C.B-17 SCID mice 8–12 weeks of age that had received 975 or 375 cGy total body irradiation, respectively.

Quantification of Systemic tsFlk-1 and VEGF Expression.

Conditioned medium from NXS2 cells that had been transduced directly with the MSCV-tsFlk-1-I-GFP retroviral vector was used as the source for tsFlk-1 protein. 6-Histidine-labeled tsFlk-1 was purified from the medium using Ni2+-NTA-beads (Qiagen, Santa Clarita, CA) as described previously (17). Purity of protein recovery was evaluated by Coomassie Blue staining, and tsFlk-1 concentration was determined by the Bio-Rad Protein Assay using BSA as a standard. Western blot analysis was performed as described previously (17) using a goat antimouse VEGFR-2 (R & D Systems) primary antibody. The level of tsFlk-1 expression in the serum was quantified using an ELISA assay. Ninety-six-well plates (Nunc) were incubated overnight at 4°C with the anti-VEGFR-2 (R & D Systems) antibody (2.5 μg/ml) dissolved in 0.1 m sodium bicarbonate buffer (pH 10.6). The next day, the wells were washed with PBS and blocked with 3% BSA/2% sucrose in PBS at room temperature for 30 min. Then serum samples (diluted 1:500 with 1% BSA in PBS) were added and incubated at room temperature for 1 h. Then the wells were washed with 1% BSA/0.05% Tween 20 in PBS and biotinylated anti-VEGFR-2 (50 μg/ml 1% BSA in PBS) was added. After 1 h, streptavidin-conjugated alkaline phosphatase was added, and subsequent detection was achieved using a chromogenic substrate from CytImmune Sciences (College Park, MD) according to the manufacturer’s instructions. Levels were quantified by comparing sample absorbance values to a standard curve generated by recombinant tsFlk-1 protein dissolved in naïve mouse serum.

Quantification of systemic VEGF expression was performed on mouse serum using a murine VEGF ELISA kit (Quantikine; R & D Systems) according to the manufacturer’s instructions.

Analysis of Endothelial Cell Precursor Frequency.

FACS analysis was performed to compare the frequency of endothelial cell precursors in the tsFlk-1- and the GFP-transplanted cohorts of mice. Peripheral blood was collected in EDTA from four mice in each group. After RBC lysis, primary staining was performed with either a monoclonal antibody against mouse VE-cadherin, CD144, (PharMingen, San Diego, CA), which was used as an endothelial lineage marker, or rat IgG2aκ (PharMingen), which served as an isotype control. These were diluted 1:2 and 1:400 respectively. A PE-conjugated antirat antibody (1:10; Caltag Laboratory, Burlingame, CA) was used as the secondary antibody. Analysis of CD144-positive cells was then performed by FACS after gating for monocyte size.

Serum Inhibition of VEGF-stimulated HUVEC Migration.

Endothelial cell migration assays were performed as described previously (17). Endothelial Growth Medium supplemented with VEGF (10 ng/ml; R & D Systems) along with serum (5%) from tsFlk-1- or GFP-transplanted mice was placed in the bottom wells, and HUVECs (passage 5) were added to the upper chambers. The plates were incubated for 6 h at 37°C with 5% CO2 to allow the cells to migrate. Nonmigrated cells on the upper surface of the filter were removed by wiping with a cotton swab. Then the cells were fixed with 10% formalin, stained with Harris’ hematoxylin, and, after washing, the migrated cells were counted. The assays were run in triplicate.

Murine Tumor Models.

Heterotopic neuroblastomas were established in syngeneic female A/J mice by s.c. injection of 106 NXS2 tumor cells in 200 μl of PBS. Wilms’ tumor xenografts were established by injecting 2 × 106 SK-NEP-1 cells s.c. in C.B-17 SCID mice. Tumor measurements were performed in two dimensions with calipers twice weekly, and volumes were calculated as width (2) × length × 0.5. All mice had undergone bone marrow transplantation previously, with cells that had been modified to express either tsFlk-1 together with GFP, or GFP alone. These experiments were performed in accordance with a protocol approved by the Institutional Animal Care and Use Committee of St. Jude Children’s Research Hospital.

Immunohistochemistry.

Tumors were excised from experimental mice, placed in Tissue Freezing Medium (Triangle Biomedical Sciences, Durham, NC) and snap-frozen in liquid nitrogen. Six-μm sections were fixed with cold acetone, rinsed twice with PBS, and blocked with 3% hydrogen peroxide in PBS for 12 min. The samples were washed three times with PBS and incubated for 10 min at room temperature with a protein blocking solution consisting of PBS (pH 7.5) and containing 5% normal horse serum and 1% normal goat serum. Excess blocking solution was drained, and the samples were incubated for 18 h at 4°C with a 1:100 dilution of monoclonal rat anti-CD31 antibody (PharMingen). Then the samples were rinsed four times with PBS and incubated for 60 min at room temperature with the appropriate dilution of peroxidase-conjugated antirat IgG. The slides were rinsed with PBS and incubated with diaminobenzidine (Research Genetics, Huntsville, AL). Then the sections were rinsed three times with distilled water and counterstained.

Stained tumor sections were scanned at low power, and the areas of greatest CD31-positive density were chosen for quantification of intratumoral vessel density. The number of individual brown-staining endothelial cells or clusters in these areas were then counted at ×400. Microvessel density counts, in two areas/section, were determined by two independent, blinded observers.

Fluorescent staining was performed as described above, except that slides were fixed in 4% paraformaldehyde, and blocking was done with 10% normal goat serum/0.1% BSA but without hydrogen peroxide. The rat antimouse CD34 antibody (PharMingen) was used at a 1:200 dilution, as was the rabbit anti-GFP antibody (Molecular Probe, Eugene, OR). Immunohistochemical detection of GFP in tumor sections seemed to be more sensitive and equally specific as compared with direct tissue fluorescence in demonstrating GFP expression. Controls in which only the secondary antibodies were added were routinely performed to exclude nonspecific cross-reactivity of the secondary antibodies. Alexa Fluor goat antirabbit (Molecular Probe) and Cy3 donkey antirat (Jackson ImmunoResearch, West Grove, PA) secondary antibodies were also used at 1:200 dilutions.

Quantification of Intratumoral tsFlk-1 Expression.

Tumors were excised from tsFlk-1- and GFP-transplanted mice and protein lysates made by homogenizing tumor specimens using a Dounce (Kontes, Vineland, NJ) homogenizer in 3 ml of lysis buffer [25 mm Tris-HCl, 150 mm HCl, 0.5% NP40, 0.5% sodium deoxycholate, 0.2% SDS, 1.0 mg of Pefabloc SC (Boehringer Mannheim, Indianapolis, IN), and 1 protease inhibitor tablet (Boehringer Mannheim)]/1.0 g of tissue. The homogenates were then incubated on ice for 30 min and centrifuged at 10,000 × g for 10 min at 4°C. The supernatants were then recentrifuged, collected, and frozen at −70°C for later use. Total protein levels were determined for each specimen using the Bradford Assay (Bio-Rad). Intratumoral levels of tsFlk-1 were quantified by ELISA as described previously.

Statistical Analyses.

Results are reported as means ± SE. Student’s t test was used to analyze statistical differences among groups in the HUVEC migration assays, microvessel density determination, VEGF expression levels, and between-time-points tumor-growth curves. A P < 0.05 was considered to be statistically significant.

Transplantation of Mice with Gene-modified Bone Marrow Cells.

Retroviral vector producer cells that generated replication incompetent ecotropic retrovirus were made based on the MSCV-tsFlk-1-I-GFP plasmid using GP+E86 packaging cells. The tsFlk-1 protein encoded for by this expression cassette had been shown previously to be a functional inhibitor of endothelial cell activation in vitro but to have no direct effect on tumor cell growth in vitro(17, 18). Additional producer cells were made that generated GFP-only retroviral vectors to serve as controls. Each of these producer cell lines generated replication-defective retrovirus at a titer of ∼5 × 105 infectious units/ml. Mus dunni (American Type Culture Collection) coculture was performed to confirm the absence of replication-competent retrovirus. Ex vivo transduction of bone marrow-derived cells was performed by coculture of these cells with the retroviral vector-producer cells. After transplantation, successful engraftment was confirmed by the presence of a normal complete blood count for A/J mice transduced with each of the vectors. The complete blood count remained normal for each group of transplanted mice (Table 1), some of which were followed for up to 1 year after bone marrow transplantation. Although there was some concern that the process of engraftment after bone marrow transplantation might be affected by angiogenesis inhibition, expression of tsFlk-1 by the bone marrow-derived cells did not preclude engraftment of these cells. There was also no long-term selection against the hematopoietic cells that had been modified to express Flk-1. The percentage of cells that were GFP-positive remained stable for each of the transplant groups in each of the hematopoietic lineages over the course of 6 months (Table 1). That the percentage of GFP-positive WBCs in the Flk-1-I-GFP mice was less than in the control mice probably reflects less efficient transduction of these cells rather than an inhibitory effect on WBC maturation by tsFlk-1, because the absolute WBC counts were equivalent between the two groups and remained stable, and the transduction of RBCs and platelets was also less efficient for the tsFlk-1-transplanted group.

Expression of a Functional Endothelial Cell Activation Inhibitor.

The level of transgene expression in the sera of tsFlk-1 transplanted A/J mice was determined by ELISA. Serum levels of Flk-1 (mean, 12.3 μg/ml ± 0.7) peaked within 6–8 weeks after transplant (Fig. 1) and remained stable for >1 year (data not shown). A similar level of tsFlk-1 expression was detected in the C.B-17 SCID mice transplanted with MSCV-Flk-1-I-GFP-transduced bone marrow.

Western blot analysis of sera from tsFlk-1-I-GFP-transplanted mice confirmed that a protein of the correct molecular weight was being expressed (Fig. 2,A). It also demonstrated that a small amount of soluble Flk-1 could be detected in sera from the control mice. That the tsFlk-1 protein being expressed in the tsFlk-1-I-GFP-transplanted mice maintained its endothelial cell inhibitory activity was confirmed by testing the ability of sera from these mice to inhibit endothelial cell migration. VEGF-stimulated HUVEC migration was inhibited by sera from mice transplanted with bone marrow-derived cells expressing the tsFlk-1 protein but not by sera from GFP-transplanted control mice (Fig. 2 B).

Circulating Endothelial Cell Precursor Frequency and Systemic VEGF Levels.

FACS analysis of peripheral blood from transplanted mice for the endothelial cell surface marker VE-cadherin (CD144) was performed to determine whether the expression of tsFlk-1 from transduced bone marrow cells had any effect on the frequency of circulating endothelial precursor cells. This analysis revealed that the tsFlk-1-transplanted mice actually have had a slightly higher frequency of circulating endothelial cell precursors than the control GFP-transplanted mice (0.78% ± 0.23 versus 0.42% ± 0.18), although this difference did not reach statistical significance. Similarly, the serum VEGF levels were higher in the tsFlk-1-transplanted mice than the GFP-transplanted control mice (43.11 pg/ml ± 10.82 versus 25.92 pg/ml ± 7.19; P < 0.02).

Inhibition of in Vivo Tumor Growth.

Next we sought to determine what effect this systemic state of angiogenesis inhibition would have on in vivo tumor growth. First, we evaluated the effect in a localized murine neuroblastoma model. NXS2 cells (1 × 106) were implanted in the s.c. space over the flank of syngeneic, immunocompetent A/J mice. These mice had been transplanted two months before with bone marrow cells transduced with either MSCV-Flk-1-I-GFP or MSCV-I-GFP. Ten mice in each group were studied. Tumor growth over time in this heterotopic location is shown graphically in Fig. 3,A. Tumors grown in mice transplanted with truncated Flk-1-expressing bone marrow cells were 50% smaller than those which developed in GFP-control-transplanted mice 25 days after tumor cell injection (mean volume, 602.6 mm3 ± 211.2 versus 1175.0 mm3 ± 142.1; P < 0.02 at day 25 only). The mean intratumoral tsFlk-1 level in these tumors was 0.75 ng ± 0.05/μg total protein as compared with 0.08 ± 0.02 ng/μg total protein within tumors in the GFP-transplanted mice. Immunohistochemical evaluation of tumor vascularity was performed to confirm that tumor growth restriction in the tsFlk-1-bone marrow-transplanted mice was attributable, at least in part, to the inhibition of tumor-induced angiogenesis. A greater density of capillary vessels was seen in tumors that grew in control GFP-transplanted mice as compared with those tumors that grew in tsFlk-1-transplanted mice (mean, 25.92 ± 7.19 v. 20.75 ± 7.44; P < 0.1) although this difference did not reach statistical significance. Representative photomicrographs are shown in Fig. 4. An even more profound inhibition of tumor growth was observed in the heterotopic Wilms’ tumor xenograft model (Fig. 3 B). Seven mice in each group were studied. In this model, tumors grown in mice transplanted with tsFlk-1-expressing bone marrow cells were less than 10% the size of those which developed in GFP-control transplanted mice 42 days following tumor cell injection (mean volume = 61.3 mm3 ± 24.0 versus 716.7 mm3 ± 156.6, P < 0.001 for all time points after day 21).

GFP Analysis of Tumor Vasculature.

Tumor growth and concordant neovascularization in mice whose bone marrow cells have been modified to express the marker protein GFP afforded us the opportunity to test the hypothesis that tumor-induced angiogenesis is supported by bone marrow-derived endothelial cell precursors. Tumors grown in control GFP-bone marrow transplanted mice were evaluated by immunofluorescence for colocalization of endothelial cell markers and GFP in tumor neovasculature. Two different antiendothelial antibodies (anti-CD31 and anti-CD34), with slightly different cross-reactivity profiles, were used to identify the endothelial cells within the tumor specimens. The vascular staining pattern within the tumors, using each of these antibodies, was similar. Approximately 5% of these CD31/CD34-positive cells, usually those at the tumor periphery, also expressed GFP (Fig. 5). Other transduced hematopoietic bone marrow-derived cells within the tumor, in addition to the endothelial cells, are also seen expressing GFP (Fig. 5 A). A similar pattern of colocalization was observed in the smaller tumors that grew in the tsFlk-1-transplanted mice (data not shown). The demonstration that GFP expression localizes with expression of endothelial cell markers suggests that these endothelial cells did, in fact, arise from the GFP-modified bone marrow-derived cells used for transplantation.

Inhibition of tumor-associated angiogenesis, in which the activated endothelial cells, rather than the tumor cells, are targeted, has become an increasingly popular anticancer approach. This strategy is particularly appealing for several reasons. (a) It is likely that most, if not all, types of cancer are dependent on angiogenesis, thereby providing a common target in the treatment of widely heterogeneous tumor types. (b) Endothelial cells are normal cells with a low intrinsic mutation rate and therefore are less likely to acquire a drug-resistant phenotype than the genetically unstable tumor cells. The development of drug resistant clones within a population of chemosensitive tumor cells is often the event that ultimately causes the failure of traditional anticancer approaches. And (c), tumor-activated endothelial cells can be selectively targeted because, with the exception of endothelial cells activated for new vessel formation required in wound healing and reproduction, nearly all other host endothelial cells are quiescent.

Gene therapy-mediated delivery of angiogenesis inhibitors is an attractive alternative means for providing long-term expression of these therapeutic proteins. In this approach, host cells are engineered to make the antiangiogenic protein in vivo on a continuous basis, thereby obviating the need for daily administration of recombinant protein. A number of different host tissues are currently being used as targets for gene therapy-mediated delivery of therapeutic proteins. These have included, most often, skeletal muscle and liver, chosen because of their relatively low mitotic activity. This decreases the likelihood that an episomal transgene will be lost during cell division, thus resulting in the potential for long-term expression. Two common methods for gene-transfer include liposome or adenoviral-mediated delivery. However, the transfer of naked DNA is typically an inefficient process, whereas adenoviral-mediated gene transfer is complicated by a host immune response to transduced target cells, resulting in transient, albeit high, transgene expression. Retroviral-mediated transgene integration into self-renewing stem cells is an attractive alternative for achieving long-term transgene expression.

Bone marrow-derived stem cells are among the most accessible self-renewing cells, and protocols exist for the efficient, retroviral-mediated transduction of these cells in mice. In addition, we hypothesized that these stem cells might be a source of endothelial cell precursors recruited for tumor-induced neovascularization. Although the dependence of tumor growth on supportive new blood vessel formation is now widely accepted, the origin of the endothelial cells that comprise tumor neovasculature is less certain. It had been assumed that the additional endothelial cells required to construct these new vessels come from the division and proliferation of local endothelial cells. However, the existence of bone marrow-derived endothelial cell precursors, or hemangioblasts, has recently been described (21, 22, 23, 24). These endothelial progenitor and stem cells have been detected in the peripheral blood and have the capacity to differentiate into endothelial cells in vitro and in vivo(21, 22, 23, 24). It has been suggested that endothelial cells incorporated into sites of neovascularization, including tumor-induced new blood vessels, may, in fact, be derived from these precursor cells of bone marrow origin (12, 25, 26, 27). These progenitors can be mobilized from the bone marrow in response to ischemia and VEGF (21, 26), and can be recruited in the colonization of endothelial flow surfaces of vascular prostheses and the neovasculature associated with cornea micropocket surgery (22, 25). Our study has confirmed that these circulating bone marrow-derived endothelial cell precursors are recruited to tumor-induced neovascularization. No GFP expression was detected in the parenchyma or vasculature of other tissues within the transplanted mice, including kidney, liver, adrenal, spleen, lung, aorta, and heart (data not shown), however, despite recent evidence to suggest to bone marrow cells include precursors to a wide variety of tissue types (28). This is probably because, in our model system, there was no impetus for accelerated turnover or regeneration of these tissues.

This finding that circulating bone marrow-derived endothelial cells are recruited to tumor neovasculature provides additional rationale for the use of bone-marrow stem cells as the target in an antiangiogenic gene therapy-mediated anticancer strategy. Because these gene-modified endothelial cell precursors are recruited by tumor-elaborated factors to sites of tumor growth, a higher local milieu of angiogenesis inhibitors is likely established at the desired target sites. This observation also has important therapeutic implications for the specific targeting of other classes of anticancer agents to the tumor microenvironment. The relative contributions of bone marrow-derived precursor cells and endothelial cells recruited from local, established blood vessels to tumor neovascularization are uncertain. Although 50–80% of hematopoietic cells expressed GFP after transduction and subsequent engraftment, a smaller percentage of endothelial cells of the tumor neovasculature expressed detectable GFP by fluorescence immunohistochemistry. This may be attributable to a number of factors, including the inability to detect low-level GFP expression, a lower transduction efficiency of endothelial cell precursors, or the fact that the non-GFP-expressing endothelial cells were derived from local vasculature. The later explanation would support the hypothesis that both vasculogenesis and angiogenesis contribute to postnatal, tumor-induced neovascularization. Nevertheless, the finding of GFP expression in some of the endothelial cells in the tumor neovasculature of our transplanted mice confirms the origin of these cells from bone marrow precursors.

This study marks the first time in which long-term expression of high levels of a functional angiogenesis inhibitor has been established by a gene therapy approach. Serum levels of the transgene peaked within 6–8 weeks after transplant and remained stable for >1 year. This finding confirmed the following. (a) Bone marrow-derived stem cells had been successfully transduced. (b) Expression of this angiogenesis inhibitor did not adversely affect bone marrow engraftment. (c) Despite recent observations that have suggested a role for Flk-1 in murine embryonic hematoangiogenesis (29), generation of mature hematopoietic cells (or endothelial cells) after bone marrow transplantation was not inhibited by the enforced expression of the soluble, truncated receptor from precursor cells. Perhaps this reflects differences in embryonic hematopoiesis compared with adult hematopoiesis, or it may simply be that the level of tsFlk-1 expression in the environment of the bone marrow is not high enough to effect these processes. And (d), long-term transgene expression can be generated by gene-modified transplanted bone marrow-derived cells.

It does not seem that the antiangiogenic effects seen with the approach used in this study are attributable to an inhibitory effect on endothelial cell generation or recruitment from the bone marrow but rather on the local steps required for angiogenesis at sites of tumor growth. Expression of this VEGF inhibitor did not diminish endothelial cell precursor frequency in the peripheral blood of tsFlk-1-transplanted mice. In fact, the frequency was slightly higher than in the control GFP-transplanted mice. Consistent with this was the finding of an elevated systemic level of VEGF in the tsFlk-1-transplanted mice. This may be caused by feedback resulting from inhibition of VEGF activity outside the environment of the bone marrow by tsFlk-1. In addition, although tumors grown in tsFlk-1-transplanted mice had fewer endothelial cells, the relative proportion of cells that were GFP-positive, and therefore bone marrow-derived, was approximately the same as that found in the tumors grown in the GFP-transplanted mice, also suggesting that the effect of tsFlk-1 on these endothelial cells seems to be paracrine rather than autocrine.

It is unclear whether the antiangiogenic effect of tsFlk-1 observed in these experiments is attributable, primarily, to tsFlk-1 being expressed locally either by transduced endothelial cells within the tumor neovasculature or tumor-infiltrating nonendothelial hematopoietic cells or to the high systemic levels of the angiogenesis inhibitor being expressed by all transduced bone marrow-derived cells. The inhibitory activity of HUVEC migration demonstrated by sera from mice transplanted with tsFlk-1-expressing cells suggests that a systemic state of angiogenesis inhibition had been established. However, a particularly high concentration of tsFlk-1 was also noted locally within tumors grown in these mice. Perhaps a different vector design in which tsFlk-1 expression were driven by an endothelial cell-specific promoter would help to determine whether the antiangiogenic effect was caused by local or systemic expression of the inhibitor. In addition, the use of an endothelial-specific or regulatable promoter would be useful in trying to avoid the potential side effects of chronic systemic angiogenesis inhibition. With this type of vector design, antitumor tsFlk-1 efficacy, as has been achieved with recombinant protein administration (16), but without systemic exposure, as with local administration of retroviral vector producer cells (18), might be achieved long-term.

These experiments have demonstrated that bone marrow-derived cells contribute to tumor neovasculature and when modified to express the angiogenesis inhibitor tsFlk-1 can restrict tumor growth in a syngeneic murine neuroblastoma model and a Wilms’ tumor xenograft model. One of the appeals of this strategy for the treatment of pediatric cancer is that it could be readily incorporated into existing treatment schema because bone marrow transplantation is already a part of most clinical protocols for children with high-risk malignancies. Although transduction of human hematopoietic stem cells has traditionally not been as efficient as for murine cells, there have been recent reports suggesting that this ability may be improving (30), and that a clinical trial might be feasible. It is not clear why the Wilms’ tumor xenografts were more affected by the inhibition of VEGF signaling than the murine neuroblastoma tumors. It may be that human cells in murine hosts have fewer adaptive responses to this type of therapy. Or it may reflect murine host-strain differences. Alternatively, this result may reflect a difference in angiogenic capabilities of the different tumor histologies. Perhaps, Wilms’ tumor is more susceptible to VEGF inhibition than neuroblastoma, as has been suggested recently by Kim et al.(31). This will need to be evaluated further as antiangiogenic therapy moves forward. In addition, the efficacy of this antiangiogenic strategy will need to be tested in orthotopic tumor models. Although s.c. tumor models afford the opportunity to assess antitumor efficacy continuously, it has been shown that the angiogenic factors involved in s.c. tumor growth may be different from those of orthotopic tumors (32). Improvement in the antitumor efficacy of this approach is also needed, as all mice eventually succumbed to their tumors. Either higher levels of angiogenesis inhibitor expression will be required or synergy achieved by combining this approach with a cytotoxic antitumor modality such as chemotherapy or immunotherapy. Of course, the effects of this approach on angiogenesis in such physiological processes as wound healing and reproduction would have to be evaluated.

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.

        
1

This work was supported by Grant 94-000 from the Assisi Foundation of Memphis, Grant IRG-87-008-09 from the American Cancer Society, Cancer Center Support CORE Grant P30 CA 21765, and American Lebanese Syrian Associated Charities (ALSAC).

                
3

The abbreviations used are: VEGF, vascular endothelial growth factor; Flk-1 fetal liver kinase-1; HUVEC, human umbilical vein endothelial cell; GFP, green fluorescent protein; MSCV, murine stem-cell virus; FACS, fluorescence-activated cell sorter; SCID, severe combined immunodeficient; VEGFR-2, vascular endothelial growth factor receptor-2.

Fig. 1.

Levels of Flk-1 protein detected in the serum of twelve A/J mice at time points after transplantation with MSCV-Flk-1-I-GFP-modified bone marrow cells.

Fig. 1.

Levels of Flk-1 protein detected in the serum of twelve A/J mice at time points after transplantation with MSCV-Flk-1-I-GFP-modified bone marrow cells.

Close modal
Fig. 2.

A, Western blot detection of tsFlk-1 in conditioned medium from MSCV-tsFlk-1-I-GFP and MSCV-I-GFP retroviral vector-producer cells and sera from two mice transplanted with tsFlk-1 gene-modified cells and one with GFP-only gene-modified cells. B, effect of sera from mice transplanted with tsFlk-1-expressing bone marrow cells (Flk-BMT) on VEGF-stimulated HUVEC migration, as compared with sera from nontransplanted mice (naïve) and control-transplanted mice (GFP-BMT). Sera were pooled for testing from four mice in each group. Also shown are the effects of diluting the Flk-BMT sera with sera from GFP-BMT mice at ratios of 1:2 and 2:1 and the effect of medium alone (without mouse serum) with and without VEGF.

Fig. 2.

A, Western blot detection of tsFlk-1 in conditioned medium from MSCV-tsFlk-1-I-GFP and MSCV-I-GFP retroviral vector-producer cells and sera from two mice transplanted with tsFlk-1 gene-modified cells and one with GFP-only gene-modified cells. B, effect of sera from mice transplanted with tsFlk-1-expressing bone marrow cells (Flk-BMT) on VEGF-stimulated HUVEC migration, as compared with sera from nontransplanted mice (naïve) and control-transplanted mice (GFP-BMT). Sera were pooled for testing from four mice in each group. Also shown are the effects of diluting the Flk-BMT sera with sera from GFP-BMT mice at ratios of 1:2 and 2:1 and the effect of medium alone (without mouse serum) with and without VEGF.

Close modal
Fig. 3.

A, growth of 106 NXS2 tumor cells following s.c. injection into mice transplanted with tsFlk-1 gene-modified cells (○) or GFP-only gene-modified cells (•; n = 10 for each group; P < 0.02). B, growth of 2 × 106 SK-NEP-1 tumor cells after s.c. injection into mice transplanted with tsFlk-1 gene-modified cells (○) or GFP-only gene-modified cells (•; n = 7 for each group; P < 0.001).

Fig. 3.

A, growth of 106 NXS2 tumor cells following s.c. injection into mice transplanted with tsFlk-1 gene-modified cells (○) or GFP-only gene-modified cells (•; n = 10 for each group; P < 0.02). B, growth of 2 × 106 SK-NEP-1 tumor cells after s.c. injection into mice transplanted with tsFlk-1 gene-modified cells (○) or GFP-only gene-modified cells (•; n = 7 for each group; P < 0.001).

Close modal
Fig. 4.

Comparison of the vascularity of NXS2 tumors grown s.c. in mice transplanted with GFP-only gene-modified cells (A) or tsFlk-1 + GFP gene-modified cells (B), as assessed by immunohistochemical staining with an anti-CD31 antibody. Original magnification, ×10.

Fig. 4.

Comparison of the vascularity of NXS2 tumors grown s.c. in mice transplanted with GFP-only gene-modified cells (A) or tsFlk-1 + GFP gene-modified cells (B), as assessed by immunohistochemical staining with an anti-CD31 antibody. Original magnification, ×10.

Close modal
Fig. 5.

Immunofluorescence staining of a tumor excised from a mouse transplanted with GFP gene-modified cells and subsequently challenged with NXS2 cells. Staining with an anti-GFP marker is shown in A (original magnification, ×20), D (×63), and G (×40). Staining of the same sections for the endothelial cell markers is shown in B (CD31), E (CD31), and H (CD34). Double staining (yellow) is shown in C, F, and I. The final two panels show anti-CD31 (J) and anti-GFP (K) staining of a tumor grown in a nontransplanted control mouse.

Fig. 5.

Immunofluorescence staining of a tumor excised from a mouse transplanted with GFP gene-modified cells and subsequently challenged with NXS2 cells. Staining with an anti-GFP marker is shown in A (original magnification, ×20), D (×63), and G (×40). Staining of the same sections for the endothelial cell markers is shown in B (CD31), E (CD31), and H (CD34). Double staining (yellow) is shown in C, F, and I. The final two panels show anti-CD31 (J) and anti-GFP (K) staining of a tumor grown in a nontransplanted control mouse.

Close modal
Table 1

Hematopoietic reconstitution after bone marrow transplantation of A/J mice with gene-modified cellsa

Retroviral vectorCell type2 mo6 mo
MSCV-Flk-1-I-GFP Red blood cellsb (%) 43 ± 1.1 44 ± 1.2 
 GFP (%) 48 ± 0.6 48 ± 0.3 
 White blood cellsc (K/μl) 8.1 ± 0.3 7.5 ± 0.4 
 GFP (%) 57 ± 0.9 57 ± 1.0 
 Platelets (K/μl) 1111 ± 69 1260 ± 45 
 GFP (%) 55 ± 0.3 53 ± 0.4 
MSCV-I-GFP Red blood cellsb (%) 40 ± 0.8 45 ± 0.8 
 GFP (%) 54 ± 0.9 52 ± 0.5 
 White blood cellsc (K/μl) 7.4 ± 0.4 7.7 ± 0.6 
 GFP (%) 79 ± 1.3 80 ± 0.6 
 Platelets (K/μl) 1054 ± 69 1246 ± 56 
 GFP (%) 62 ± 0.4 58 ± 0.2 
Retroviral vectorCell type2 mo6 mo
MSCV-Flk-1-I-GFP Red blood cellsb (%) 43 ± 1.1 44 ± 1.2 
 GFP (%) 48 ± 0.6 48 ± 0.3 
 White blood cellsc (K/μl) 8.1 ± 0.3 7.5 ± 0.4 
 GFP (%) 57 ± 0.9 57 ± 1.0 
 Platelets (K/μl) 1111 ± 69 1260 ± 45 
 GFP (%) 55 ± 0.3 53 ± 0.4 
MSCV-I-GFP Red blood cellsb (%) 40 ± 0.8 45 ± 0.8 
 GFP (%) 54 ± 0.9 52 ± 0.5 
 White blood cellsc (K/μl) 7.4 ± 0.4 7.7 ± 0.6 
 GFP (%) 79 ± 1.3 80 ± 0.6 
 Platelets (K/μl) 1054 ± 69 1246 ± 56 
 GFP (%) 62 ± 0.4 58 ± 0.2 
a

n = 10 mice in each group.

b

Expressed as hematocrit.

c

No difference in cell number or percentage of GFP was observed when the white blood cells were analyzed further for granulocytes and lymphocytes.

We thank Dorothy Bush, Bonnie Greer, and Adriana Nance for their assistance with immunohistochemistry and Dr. Richard Ashmun, Ed Wingfield, and Anne-Marie Hamilton-Easton for their assistance with FACS. We also thank Drs. Brian Sorrentino and Stephen Shochat for their critical review of this manuscript.

1
Folkman J. The role of angiogenesis in tumor growth.
Semin. Cancer Biol.
,
3
:
65
-71,  
1992
.
2
Bicknell R., Harris A. L. Mechanisms and therapeutic implications of angiogenesis.
Curr. Opin. Oncol.
,
8
:
60
-65,  
1996
.
3
Pluda J. M. Tumor-associated angiogenesis: mechanisms, clinical implications, and therapeutic strategies.
Semin. Oncol.
,
24
:
203
-218,  
1997
.
4
Folkman J., Ingber D. Inhibition of angiogenesis.
Semin. Cancer Biol.
,
3
:
89
-96,  
1992
.
5
Bicknell R., Harris A. L. Anticancer strategies involving the vasculature: vascular targeting and the inhibition of angiogenesis.
Semin. Cancer Biol.
,
3
:
399
-407,  
1992
.
6
Scott P. A. E., Harris A. L. Current approaches to targeting cancer using antiangiogenesis therapies.
Cancer Treat. Rev.
,
20
:
393
-412,  
1994
.
7
Folkman J. Antiangiogenic gene therapy.
Proc. Natl. Acad. Sci. USA
,
95
:
9064
-9066,  
1998
.
8
Kong H. L., Crystal R. G. Gene therapy for tumor antiangiogenesis.
J. Natl. Cancer Inst.
,
90
:
273
-286,  
1998
.
9
Feldman A. L., Libutti S. K. Progress in antiangiogenic gene therapy.
Cancer (Phila.)
,
89
:
1181
-1194,  
2000
.
10
Klement G., Baruchel S., Rak J., Man S., Clark K., Hicklin D. J., Bohlen P., Kerbel R. S. Continuous low-dose therapy with vinblastine and VEGF receptor-2 antibody induces sustained tumor regression without overt toxicity.
J. Clin. Investig.
,
105
:
R15
-R24,  
2000
.
11
Spencer H. T., Sleep S. E., Rehg J. E., Blakley R. L., Sorrentino B. P. A gene transfer strategy for making bone marrow cells resistant to trimetrexate.
Blood
,
87
:
2579
-2587,  
1996
.
12
Asahara T., Masuda H., Takahashi T., Kalka C., Pastore C., Silver M., Kearne M., Magner M., Isner J. M. Bone marrow origin of endothelial progenitor cells responsible for postnatal vasculogenesis in physiological and pathological neovascularization.
Circ. Res.
,
85
:
221
-228,  
1999
.
13
Dvorak H. F., Brown L. F., Detmar M., Dvorak A. M. Vascular permeability factor/vascular hyperpermeability, and angiogenesis.
Am. J. Pathol.
,
146
:
1029
-1039,  
1995
.
14
Millauer B., Shawver L. K., Plate H. K., Risau W., Ullrich A. Glioblastoma growth inhibited in vivo by a dominant-negative Flk-1 mutant.
Nature (Lond.)
,
367
:
576
-579,  
1994
.
15
Millauer B., Longhi M. P., Plate K. H., Shawver L. K., Risau W., Ulrich A., Strawn L. M. Dominant-negative inhibition of Flk-1 suppresses the growth of many tumor types in vivo.
Cancer Res.
,
56
:
1615
-1620,  
1996
.
16
Lin P., Sankar S., Shan S., Dewhirst M. W., Polverini P. J., Quinn T. Q., Peters K. G. Inhibition of tumor growth by targeting tumor endothelium using a soluble vascular endothelial growth factor receptor.
Cell Growth Differ.
,
9
:
49
-58,  
1998
.
17
Davidoff A. M., Leary M. A., Ng C. Y. C., Vanin E. F. Gene therapy mediated delivery of the angiogenesis inhibitor FLK-1 results in inhibition of neuroblastoma growth in vivo.
J. Pediatr. Surg.
,
36
:
30
-36,  
2000
.
18
Davidoff A. M., Leary M. A., Ng C. Y. C., Vanin E. F. Gene therapy-mediated angiogenesis inhibition restricts neuroblastoma growth in vivo.
Med. Pediatr. Oncol.
,
35
:
638
-640,  
2000
.
19
Lode H. N., Xiang R., Varki N. M., Dolman C. S., Gillies S. D., Reisfeld R. A. Targeted interleukin-2 therapy for spontaneous neuroblastoma metastases to bone marrow.
J. Natl. Cancer Inst.
,
89
:
1586
-1594,  
1997
.
20
Dubridge R. B., Tang P., Hsia H. C., Leong P. M., Miller J. H., Calos M. P. Analysis of mutation in human cells by using an Epstein Barr virus shuttle system.
Mol. Cell. Biol.
,
7
:
379
-387,  
1987
.
21
Asahara T., Murohara T., Sullivan A., Silver M., van der Zee R., Li T., Witzenbichler B., Schatteman G., Isner J. M. Isolation of putative progenitor endothelial cells for angiogenesis.
Science (Wash. DC)
,
275
:
964
-967,  
1997
.
22
Shi Q., Rafii S., Wu M. H., Wijelath E. S., Yu C., Ishida A., Fujita Y., Kothari S., Mohle R., Sauvage L. R., Moore M. A. S., Storb R. F., Hammond W. P. Evidence for circulating bone marrow-derived endothelial cells.
Blood
,
92
:
632
-367,  
1998
.
23
Choi K., Kennedy M., Kazarov A., Papadimitriou J. C., Keller G. A common precursor for hematopoietic and endothelial cells.
Development
,
125
:
725
-732,  
1998
.
24
Lin Y., Weisdorf D., Solovey A., Hebbel R. Origins of circulating endothelial cells and endothelial outgrowth from blood.
J. Clin. Investig.
,
105
:
71
-77,  
2000
.
25
Takahashi T., Kalka C., Masuda H., Chen D., Silver M., Kearney M., Magner M., Isner J. M., Asahara T. Ischemia- and cytokine-induced mobilization of bone marrow-derived endothelial progenitor cells for neovascularization.
Nat. Med.
,
5
:
434
-438,  
1999
.
26
Asahara T., Takahashi T., Masuda H., Kalka C., Chen D., Iwaguro H., Inai ., Silver M., Isner J. M. VEGF contributes to postnatal neovascularization by mobilizing bone marrow-derived endothelial progenitor cells.
EMBO
,
18
:
3964
-3972,  
1999
.
27
Crosby J. R., Kaminski W. E., Schatteman G., Martin P. J., Raines E. W., Seifert R. A., Bowden-Pope D. F. Endothelial cells of hematopoietic origin make a significant contribution to adult vessel formation.
Circ. Res.
,
87
:
728
-730,  
2000
.
28
Krause D. S., Theise N. D., Collector M. I., Henegariu O., Hwang S., Gardner R., Neutzel S., Sharkis S. J. Multi-organ, multi-lineage engraftment by a single bone marrow-derived stem cell.
Cell
,
105
:
369
-377,  
2001
.
29
Ziebler B., Valtieri M., Porada G. A., De Maria R., Muller R., Masella B., Gabbianelli M., Casella I., Pelosi E., Bock T., Zanjani E. D., Peschle C. KDR receptor: a key marker defining hematopoietic stem cells.
Science (Wash. DC)
,
285
:
1553
-1558,  
1999
.
30
Kelly P. F., Vandergriff J., Nathwani A., Nienhuis A. W., Vanin E. F. Highly efficient gene transfer into cord blood nonobese diabetic/severe combined immunodeficiency repopulating cells by oncoretroviral vector particles pseudotyped with the feline endogenous retrovirus (RD114) envelop protein.
Blood
,
96
:
1206
-1214,  
2000
.
31
Kim E., Moore J., Huang J., Soffer S., Manley C. A., O’Toole K., Middlesworth W., Stolar C. J., Kandel J. J., Yamashiro D. J. All angiogenesis is not the same: distinct patterns of response to antiangiogenic therapy in experimental neuroblastoma and Wilms’ tumor.
J. Pediatr. Surg.
,
36
:
287
-290,  
2001
.
32
Yancopoulos G. D., Davis S., Gale N. W., Rudge J. S., Wiegand S. J., Holash J. Vascular-specific growth factors and blood vessel formation.
Nature (Lond.)
,
407
:
242
-248,  
2000
.