Cell-based delivery of therapeutic viruses has potential advantages over systemic viral administration, including attenuated neutralization and improved viral targeting. One of the exciting new areas of investigation is the potential ability of endothelial-lineage cells to deliver genes to the areas of neovascularization. In the present study, we compared two types of endothelial-lineage cells [outgrowth endothelial cells (OECs) and culture-modified mononuclear cells (CMMCs), also known as “endothelial progenitor cells”] for their ability to be infected with adenovirus and to home to the areas of neovascularization. Both cell types were isolated from peripheral blood of healthy human donors and expanded in culture. We demonstrate that OECs are more infectable and home better to tumors expressing VEGF on systemic administration. Furthermore, we used an adenoviral/retroviral chimeric system to convert OECs to retrovirus-producing cells. When injected systemically into tumor-bearing mice, OECs retain their ability to produce retrovirus and infect surrounding tumor cells. Our data demonstrate that OECs could be efficient carriers for viral delivery to areas of tumor neovascularization.
- stem cells
- gene therapy
- viral vectors
- cell carriers
systemic viral delivery of therapeutic genes to tumors has several limitations: achievable viral titers at the target site are, for the most part, insufficient, and viruses adhere nonspecifically to endothelium and are neutralized by the immune system. These limitations, together, impair the ability of systemically delivered viruses to successfully target tumors and metastases (26). One potential solution for improving tumor targeting is to engineer cells as “viral factories,” thereby combining cell tropism for tumors with the amplification effect provided by viruses (14). We previously showed that T cells, when converted to retrovirus-producing cells, have the ability to home to tumors, release retrovirus in a tumor-dependent manner, and achieve therapeutic efficacy (6, 7).
Tumors critically depend on developing neovasculature to grow and metastasize (5). Bone marrow-derived endothelial-lineage cells can be found in tumors after bone marrow transplantation (2) and are an essential component for tumor development in some murine models (17). Additionally, endothelial-lineage cells derived in culture from multipotent adult progenitor cells (24) and from CD34-positive hematopoietic stem cells (11) can be found incorporated in the tumor vasculature. Therefore, endothelial-lineage cells are attractive candidates for virus delivery to tumors, because, if incorporated into neovasculature, virus release will be directly proximal to the tumor cells.
Recently, it became clear that at least two types of endothelial-lineage cells can be obtained by culture of peripheral blood mononuclear cells. Culture-modified mononuclear cells (CMMCs, also called “endothelial progenitor cells”) are spindle-shaped cells that are almost uniformly positive for CD14 and other monocytic markers that express endothelial-lineage surface markers after short-term culture (9, 13, 15, 23, 25, 27). These cells have been extensively studied in animal models of neovascularization and, more recently, in human patients with cardiovascular disease (4). On the other hand, outgrowth endothelial cells (OECs) develop as rapidly growing, cobblestone-shaped colonies of cells that are capable of expanding >30 population doublings in long-term cultures, implying that these cells may be derived from “true” rare progenitors (10, 16, 20, 21). We previously showed that these two cell types differ in their origin and that OECs are more angiogenic than CMMCs in Matrigel angiogenesis models in vitro and in vivo, potentially because of higher endothelial nitric oxide synthase activity in OECs (12). Together, these results imply that nontransduced OECs may be more potent for treating atherosclerotic disease. However, the use of cells as autologous agents for systemic gene or virus delivery depends on 1) their ability to be transduced and, subsequently, generate virus and 2) their ability to home specifically to target tissue. Here, we evaluated the biodistribution of CMMCs and OECs after systemic administration in murine tumor models, revealing tumor specificity only with OECs, which was only marginally greater than background. However, we also demonstrate that OECs can be readily infected with adenovirus in vitro and can be converted to potent producers of retrovirus. With this in mind, we demonstrated that systemically administered OECs can successfully and specifically transfer retrovirus to tumor cells in mice, the ex vivo engineering of OECs thereby overcoming limitations of cell homing in this model and indicating proof of the concept of targeted retrovirus delivery.
MATERIALS AND METHODS
OEC and CMMC isolation and culture.
OECs and CMMCs were grown as previously described (12). Briefly, human peripheral blood mononuclear cells were obtained by Ficoll-Histopaque-1077 (Sigma-Aldrich) centrifugation from peripheral blood of healthy volunteers. Red blood cells were lysed in ammonium chloride. CMMCs were plated at 107 cells/ml, 2–3 ml/well of a six-well plate previously coated with human fibronectin (1 μg/cm2; Becton Dickinson), using EGM-2 (Clonetics, San Diego, CA) with 5% FBS without corticosteroids, as previously described (15). Medium was changed on days 4 and 7. Cells were used between days 7 and 10. OECs were plated at 107 cells/ml EGM-2 in wells of a fibronectin-coated six-well plate (2–3 ml per well). After 24 h, wells were washed once with PBS to remove nonadherent cells and fed EGM-2 daily thereafter. Colonies of rapidly proliferating monolayer cells appear between 2 and 3 wk; colonies are then chosen and expanded on fibronectin in the same medium but fed every other day.
Characterization of OECs and CMMCs.
Cells were treated with 0.05% trypsin, washed, and incubated with primary antibodies against VEGFR2 (Sigma-Aldrich), Tie-2, VE-cadherin, CD14, and CD31 (all obtained from BD Pharmingen, San Diego, CA) at 4°C for 30 min. After they were washed, the cells were incubated with secondary anti-mouse IgG-FITC antibody (BD Pharmingen) for 30 min at 4°C, washed, and fixed in 4% formaldehyde. Negative control was isotype-specific nonrelated antibody.
Homing of CMMCs and OECs to tumors.
All animal experiments were approved by the Mayo Foundation Institutional Animal Care and Use Committee. B16 cells (2 × 105) stably expressing human VEGF (B16.VEGF) were injected subcutaneously into the right flank of 6- to 8-wk-old female athymic nude mice. When tumors were 5 mm diameter, equal numbers (3 × 104–5 × 105) of labeled endothelial-lineage or control HT1080 cells were injected intravenously into the tail vein. For fluorescent labeling, cells were washed and detached using 0.05% trypsin, washed, and resuspended in 1 ml of PBS. Three microliters of CM-1,1′-dioctadecyl-3,3,3′,3′-tetramethyliodocarbocyanine (DiI; Molecular Probes, Eugene, OR) were added, and cells were incubated for 10–15 min at 37°C. Cells were washed three times in PBS and resuspended in 100 μl of PBS for intravenous injection. Tumors were harvested after 3 days, sectioned, and examined using fluorescent microscopy. Additionally, 5-μm sections were stained with biotinylated anti-human CD31 or von Willebrand factor MAb (Dako, Carpinteria, CA) and detected using the streptavidin-alkaline phosphatase kit (Dako). For detection of cell incorporation, sections of harvested tumors were incubated with fluorescein-Griffonia (Bandeiraea) simplicifolia lectin I, isolectin B4 (Vector Laboratories, Burlingame, CA), at 4°C and examined by confocal microscopy. For radioactive labeling, cells were incubated in PBS with 20 μCi of 111In oxine (Amersham, Piscataway, NJ) at room temperature for 20 min, washed three times, and resuspended in 100 μl of PBS for intravenous injection. Tumors and organs were harvested after 3 days and weighed. Radioactive counts were obtained using a gamma reader. Labeling efficiency was 20–40% in all experiments, and the total recovered dose was ∼10–20% of the injected dose. The residual radioactivity was excreted, resulting in high radioactivity of the bedding in cages.
Cell transfection, infection, and retrovirus production.
The AdenoQuest kit (Quantum Biotechnologies) was used to make Ad.LacZ and Ad.LTR.mCD2. LTR.mCD2 construct was made by replacing puromycin-resistance gene in pBabe.Puro retroviral vector (18). Ad.gag/pol and Ad.101A(env) were previously described (8). For comparing infectability of CMMCs and OECs by adenovirus, a well of a six-well plate containing CMMCs or OECs was covered with 1 ml of EGM-2 and an appropriate volume of Ad5.LacZ. Cells were incubated overnight, washed three times, and stained with 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) 48 h later. Ten ×20 fields were analyzed for the percentage of blue cells. The experiment was repeated twice.
For in vitro retrovirus production, OECs were infected with Ad.LTR.mCD2, Ad.gag/pol, and Ad.101(env) overnight. On the next day, cells were washed and incubated with fresh medium for 48 h. Supernatants were then filtered through a 45-μm filter and transferred to HT1080 cells. After 48 h, HT1080 cells were analyzed for murine CD2 (mCD2) expression by fluorescein-activated cell sorting (FACS). For in vivo virus production, infected OECs were washed three times, and 2 × 106 cells were injected intravenously into nude mice bearing 5 × 5-mm subcutaneous B16.VEGF tumors. After 3 days, livers, spleens, and tumors were harvested, and a single cell suspension was obtained and analyzed for CD2 expression by FACS.
Human CMMCs and OECs were grown from peripheral blood mononuclear cells in endothelium-specific medium, as previously described (12). Spindle-shaped CMMCs (Fig. 1, left) were obtained after short-term culture. Colonies of rapidly proliferating OECs appeared after 2–3 wk and could be expanded for >2 mo in culture (Fig. 1, right). Table 1 shows phenotype of CMMCs and OECs as determined by previous studies by our group (12) and others (16, 23, 25).
Because tumors are one of the few sites of active angiogenesis in adults, we investigated whether human CMMCs and OECs were able to localize in tumors on systemic injection. 111In oxine-labeled CMMCs, OECs, or control HT1080 fibroblasts were injected intravenously into mice bearing B16.VEGF tumors, and after 3 days organs and tumors were harvested. We chose HT1080 cells as a negative control, because these are human fibroblasts without circulating and/or tumor-homing abilities in our experience (unpublished data). Figure 2A shows the representative distribution of radioactivity per gram of tissue from 24 mice. Within the same animal, tumors contained more radioactivity per gram of tissue than the majority of examined tissues (lung, heart, muscle, blood, and brain). However, the liver, spleen, and kidney exhibited the greatest relative radioactivity, suggestive of dominant cell distribution to these organs. Interestingly, however, viable cells (human CD31-positive staining) were not detected in the liver and kidney (data not shown), indicating that the high counts in these excretory organs at least in part comprise contribution from dead cells. This is further supported by high levels of radioactivity in cage bedding and mouse skin. Because labeled cells retain 80% of time-adjusted radioactivity when cultured in vitro (data not shown), loss/redistribution of total indium in vivo likely represents sensitivity of cultured cells to intravascular sheer forces. Because of high background, we calculated relative homing of OECs and CMMCs by dividing counts per milligram of tumor by the average counts per milligram of other tissues. When relative homing between cells was compared, OECs seemed to have an advantage over CMMCs in their ability to localize in tumors (Fig. 2B), although the overall level of specificity for tumor was at best marginally better than background. Furthermore, homing of CMMCs was no better than that of control HT1080 cells. These data collectively suggest, within limitations of the xenogenic model, that homing of endothelial-lineage cells to neovascularized areas occurs, but the level of specificity is low compared with the background, particularly for CMMCs. This is consistent with recent observations of CMMC homing in nude rats with myocardial ischemia (1). Homing of OECs to B16.VEGF tumors was better than homing to tumors composed of the parental B16 cells (data not shown), supporting the notion that a degree of specificity of OECs for B16.VEGF tumors truly exists.
To visualize delivered cells, in additional experiments we labeled OECs and CMMCs with CM-DiI and injected them intravenously into nude mice bearing 5-mm subcutaneous B16.VEGF or Lewis lung carcinoma tumors. The Lewis lung carcinoma line was chosen because it has been shown to support rapid tumor neovascularization (19). After 3 days, the mice were killed and tumors were harvested. Figure 2C shows that CM-DiI-labeled cells can be detected in frozen tumor sections by fluorescent microscopy. Additionally, by staining with human-specific antibodies, the presence of human CD31- and von Willebrand factor-positive cells confirmed that the fluorescent signals in tumors correspond to injected cells and not to endogenous macrophages that may have picked up cellular debris (Fig. 2D). We also stained tumors with fluorescein-Griffonia (Bandeiraea) simplicifolia lectin I, isolectin B4, which stains rodent, but not human, endothelium. As shown in Fig. 2E, red cells contiguous with green mouse endothelium were present in tumors (arrowhead). However, more frequently, we observed perivascular fluorescent cells within tumors (arrow). On the basis of our results, OECs seem to be more promising cells for gene and viral therapy to sites of neovascularization, but additional strategies may have to be incorporated, including enhancement of targeting specificity and/or potency of delivered product.
One of the ways to amplify gene delivery to tumors is to use viruses that have selectivity for malignant cells. This can be achieved by using tumor-specific expression of key replication or therapeutic genes or, simply, by using natural properties of viruses. Retroviruses are capable of infecting only dividing cells, because intact nuclear membrane prevents retroviral entry into the nucleus. We decided to use a hybrid adenovirus-retrovirus expression system to try to deliver retrovirus with endothelial-lineage cells as viral carriers. First, we tested the ability of OECs and CMMCs to be infected with adenovirus. CMMCs and OECs were incubated overnight with the indicated multiplicity of infection of Ad5.LacZ, and the percentage of infected cells was determined after staining with X-Gal. Figure 3 shows that OECs are easily infectable with adenovirus, much more so than CMMCs. The relative resistance of CMMCs to adenovirus infection is consistent with their myelomonocytic origin, because endothelial cells are usually easily infectable with adenovirus.
Because OECs were more easily infectable with adenoviruses and expressed more homing properties to tumors than CMMCs, we decided to try to use OECs as virus carriers. To convert OECs to retrovirus-producing cells, we infected them with three separate adenoviruses encoding 1) wild-type LTR.mCD2, a murine leukemia virus-based retrovirus with the wild-type long terminal repeats (LTRs) encompassing mCD2; 2) gag/pol (8), gag and pol proteins, which are necessary for the viral life cycle; and 3) env (8), amphotropic envelope protein 101A, which is necessary for viral budding and infectivity. These triple-infected cells can produce retrovirus encoding mCD2 in culture, inasmuch as supernatant from triple-infected OECs is capable of infecting HT1080 cells (Fig. 4A). As a negative control, OECs were infected only with Ad.LTR.mCD2, and supernatant from these cells did not transfer any CD2-expressing constructs to HT1080 cells. In contrast to OECs, CMMCs were not able to produce retrovirus when infected with three adenoviruses, likely because of low adenoviral infectivity (data not shown).
Next, we wanted to test whether systemically injected OECs could produce retrovirus and transduce tumor cells in vivo. OECs were infected with three adenoviral constructs and injected into the tail vein of nude mice bearing 5-mm subcutaneous B16.VEGF tumors. After 3 days, when tumors reached ∼10 mm, mice were euthanized, and tumors were excised and homogenized into a single-cell suspension. Suspended cells were then analyzed by FACS for expression of mCD2. As shown in Fig. 4B, the liver and spleen of mice receiving triple-infected OECs did not contain any cells infected by OEC-produced retrovirus. This is likely due to inability of retrovirus to infect nondividing cells. However, in mice injected with triple-infected OECs, a subpopulation of tumor cells was positive for mCD2 expression, indicating that, despite model limitations, OECs were able to deliver retrovirus, which, in turn infected tumor cells (Fig. 4B, bottom). Tumors from mice receiving OECs infected with Ad.LTR.CD2 alone were negative for CD2, excluding the possibility that some adenovirus has been transferred to mice, together with infected OECs, or that the signal we detected reflected injected OECs only.
Cancer gene therapy has yet to produce convincing evidence for genuinely systemic gene delivery. Circulatory and immune systems rapidly and efficiently clear systemically injected DNA and viruses. The need for alternative means of delivering viruses and genes to tumors resulted in exploration of cell-based carriers as tumor site-specific viral producers (14). Initial studies focused mostly on the use of immune cells (T cells and macrophages), because homing properties of these cells have been studied extensively. However, with the advances in tumor vasculogenesis-directed therapies and stem cell research, it has became apparent that endothelial-lineage cells could be a potentially powerful tool in delivering genes to tumors. The research in this area has also been fueled by the use of the same endothelial-lineage cells in treatment of arterial injury in animal models.
Since the first discovery of circulating endothelial cell precursor (3), a number of reports have shown the potential efficacy of endothelial-lineage cells in arterial injury/repair models. However, seemingly rapid progress in clinical application research was not accompanied by true understanding of the biology and origin of these cells. More stringent criteria in defining “stem” and “progenitor” cells are needed to incorporate the field of endothelial precursor cell research with more advanced areas, such as studies of hematopoietic stem cells. In an attempt to clarify some of these issues, we have demonstrated that at least two types of endothelial-lineage cells can be obtained by culturing human peripheral blood mononuclear cells (12). Rapidly proliferating OECs have more mature endothelial phenotype and, likely, do not arise from early spindle-shaped monocytoid cells (CMMCs). This finding extends data from others who demonstrated that such cells exhibit only minimal expression of stem cell antigens and have a very low proliferative index (23). The important next step is to compare and evaluate characteristics of the two available cell populations for their use in clinical cardiovascular disease and cancer. We have shown that human OECs have an advantage over CMMCs in several useful properties for their use as viral delivery vehicles to tumors. In particular, OECs can be easily converted to retrovirus-producing cells, which are capable of infecting tumors in vivo by use of an adenovirus/retrovirus chimeric system. Despite seemingly better homing of OECs than CMMCs to tumors (Fig. 2B), our data fail to demonstrate significant homing for either of these cell types in a xenogenic model. Furthermore, even OECs localized in tumors were rarely found incorporated into blood vessels, as determined by colocalization of fluorescently labeled cells with lectin-stained murine blood vessels. This may be explained by low efficiency of human-to-mouse chimerism in building new blood vessels. Furthermore, the kinetics of B16 tumor growth may be too fast to allow significant remodeling of vasculature. Finally, endothelial-lineage cells may, in fact, have a predominant role in the hypoxic areas of the tumor, where they finish differentiation in situ and form de novo blood vessels in the process of neovasculogenesis. This differs from the neoangiogenic process in which preexisting surrounding blood vessels remodel to support surrounding ischemic tissue (22).
Regardless of low inefficient homing and incorporation, we were able to engineer OECs into virus-producing cells capable of delivering marker gene (CD2) to tumors on systemic injection. In this model system, we relied solely on the specificity of retrovirus for infecting dividing cells. Although the spleen contains dividing cells, their percentage is lower than in rapidly growing tumors, and their infectability by OEC-produced retrovirus is likely below our detection level. We foresee that OEC-mediated retroviral transfer of genes to tumor will be therapeutically valuable by replacement of the CD2 transgene by a therapeutic gene encoding a cytotoxic protein or an immunostimulatory cytokine. As we demonstrated previously (6), additional levels of specificity could be incorporated into the system, including tumor-specific promoter (e.g., tyrosinase) driving transcription of a delivered gene. This approach would avoid any potential toxicity to the liver, spleen, or kidney. Although the level of gene transfer is relatively low, we have never been able to deliver retroviral supernatant to subcutaneous tumors by systemic intravenous injection. We are improving the tumor delivery efficacy by optimizing the multiplicity of infection of triple adenoviral infection and timing of OEC injection relative to tumor growth. This is the first study that shows the ability of OECs to produce virus locally in the tumor environment in a non-bone marrow transplant model. Future studies will have to address the improvement of OEC production, homing, and viral delivery before translation of this novel technology to clinical practice for human patients.
This work was supported by National Institutes of Health Grant R-01 CA-94180 (to R. G. Vile), Cancer Research UK (to I. Hennig), and the Mayo Foundation.
The authors thank Toni L. Higgins for secretarial assistance.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2004 by the American Physiological Society