|Home | About | Journals | Submit | Contact Us | Français|
Suicide gene transfer is the most commonly used cytotoxic approach in cancer gene therapy; however, a successful suicide gene therapy depends on the generation of efficient targeted systemic gene delivery vectors. We recently reported that selective systemic delivery of suicide genes such as the herpes simplex virus thymidine kinase (HSVtk) to tumor endothelial cells via a novel targeted AAVP vector leads to suppression of tumor growth. This marked effect has been postulated to result primarily from the death of cancer cells by hypoxia following the targeted disruption of tumor blood vessels. Here we investigated whether an additional mechanism of action is involved. We show that there is a heterotypic bystander effect between endothelial cells expressing the HSVtk suicide gene and tumor cells. Treatment of co-cultures of HSVtk-transduced endothelial cells and non-HSVtk-transduced tumor cells with ganciclovir results in the death of both endothelial and tumor cells. Blocking of this effect by 18α-glycyrrhetinic acid indicates that gap junctions between endothelial and tumor cells are largely responsible for this phenomenon. Moreover, the observed bystander killing is mediated by connexin (Cx)43 and Cx26, which are expressed in both endothelial and tumor cell types. Finally, this heterotypic bystander effect is accompanied by a suppression of tumor growth in vivo that is independent of primary gene transfer into host-derived tumor vascular endothelium. These findings add an alternative non-mutually exclusive and potentially synergistic cytotoxic mechanism to cancer gene therapy based on targeted AAVP, and further support the promising role of non-malignant tumor stromal cells as therapeutic targets.
Gene therapy remains a potentially viable strategy for the treatment of human cancer. It is based on the correction of pathologic gene expression patterns (e.g. by the transfer of tumor suppressor genes) or on the delivery of cytotoxic genes that directly or indirectly kill tumor cells irrespective of its gene expression. The most widely-used approach for cytotoxic gene therapy involves the transfer of the Herpes simplex virus type I thymidine kinase (HSVtk) gene (1-3) Expression of HSVtk results in the phosphorylation of prodrug nucleoside analogues such as ganciclovir (GCV), and converts them into nucleoside analogue triphosphates. These compounds, which are incorporated into the cellular genome, inhibit DNA polymerase and cause cell death by apoptosis (4). The converted cytotoxic drug and/or toxic metabolites are able to spread from transduced cells to non-transduced cells via cellular gap junctions. This “bystander effect” may potentially overcome the requirement for all malignant cells to be transduced in order to achieve meaningful tumor regression (2, 5).
Although this approach has shown promise in vitro and in vivo, its wide application has been hampered by the lack of vectors that allow specific and efficient transduction of the target tissue after systemic administration. Consequently, poor efficiency of gene transfer potentially limits the number of vector-transduced tumor cells (6) and thus prevents effective systemic cancer gene therapy.
Given the estimates that up of 100 tumor cells are sustained by a single endothelial cell (7), vascular gene targeting might minimize or overcome this problem. Indeed, a small number of transduced cells that are accessible to the circulation could in theory mediate a much more pronounced effect that is relatively independent of gene transfer efficiency. The vasculature of a solid tumor is an attractive target for intervention because the angiogenic endothelium expresses several cell-surface receptors that are essentially absent or barely detectable in normal blood vessels (8, 9). Such receptors are suited for systemic gene therapy because they are readily accessible via the circulation and often mediate cellular internalization of targeting ligands (9-13).
Several technologies for design and production of vascular-targeted gene therapy vectors based on ligand-directed binding of the vector to endothelial receptors have been developed (14-19). We have reported a ligand-directed vector for systemic tissue-targeted gene delivery, termed adeno-associated virus/phage (AAVP; refs. 20, 21). AAVP is a genomic hybrid of adeno-associated virus type 2 (AAV-2) and of an M13 phage derivative. An established version of this vector displays the peptide ligand RGD-4C (20-22) that targets αv integrins expressed by angiogenic blood vessels (20-27). In vitro, transgene expression by targeted RGD-4C AAVP begins 48–72 h after incubation with cells and reaches a maximum level by 1 week. Transduction efficiency varies from one cell line to another depending on the expression of the target receptor. In general 10–20% of cells can be transduced in culture (20, 22, 28).
Targeted RGD-4C AAVP was also used to systemically deliver the HSVtk gene to αv integrin-positive cells in either isogenic EF43-FGF4 mouse mammary tumors (20) or nude rats bearing human sarcoma xenografts (21). EF43-FGF4 tumor cells themselves have a barely detectable expression level of αv integrin receptors that does not allow their transduction by RGD-4C AAVP; nevertheless, systemic administration of targeted RGD-4C AAVP-HSVtk to mice bearing established EF43-FGF4 tumors resulted in marked suppression of tumor growth after GCV treatment (20). Such anti-tumor effect was accompanied by extensive tumor vascular disruption caused by apoptosis of the blood vessels (20). It is not currently known, however, whether subsequent inhibition of tumor growth by RGD-4C AAVP-HSVtk plus GCV was simply a consequence of the lack of blood supply or, whether tumor cell killing was also mediated by a heterotypic bystander effect between HSVtk-transduced vascular endothelium and tumor cells.
Herein we have evaluated the hypothesis that a heterotypic bystander effect exists in this targeted system. We show that EF43-FGF4 tumor cells, which are not transduced by HSVtk, can also be eliminated both in vitro and in vivo by a vascular cell-mediated bystander effect through gap junction intercellular communication between endothelial and tumor cells.
SVEC4-10-transformed murine small vessel endothelial cells and KS1767 Kaposi's sarcoma cells were from ATCC (Manassas, VA). MDA-MB435 breast carcinoma cells were a gift from Jane Price (The University of Texas M. D. Anderson Cancer Center, Houston, TX) and 9L rat glioblastoma cells were a gift from Dr. James Basilion (Case Western Reserve University, Cleveland, OH). The EF43-FGF4 cells were derived from the EF43 BALB/c mouse mammary cell line by infection of the latter cells with a retroviral vector carrying the FGF4 oncogene, as described (20, 29). KS1767 cells were maintained in Minimal Essential Medium (MEM; Irvine Scientific, Santa Ana, CA). All other cell lines were cultured in Dulbecco's modified Eagle's Medium (DMEM; Gibco, Gaithersburg, MD). All media were supplemented with 10% FBS (Gibco), L-glutamine, and penicillin G plus streptomycin.
HSVtk was expressed in endothelial SVEC4-10 cells by transfection of a pAAV-HSVtk plasmid containing the cDNA encoding the HSVtk mutant SR39 (30). To generate the pAAV-HSVtk plasmid, we removed GFP from the pAAV-eGFP plasmid (Stratagene, La Jolla, CA) by digestion with BamHI-NotI and replaced this DNA with a BamHI-NotI fragment containing the HSVtk-SR39 (referred to in this work as HSVtk). DNA sequencing and analysis of restriction enzyme digests served to verify the correct orientation of the insert in the constructs. The plasmids were transfected into SVEC4-10 cells with the FuGENE 6 transfection reagent (Roche, IN).
For detection of Cx26 and Cx43 by immunofluorescence, cells were grown for 2-4 days to a sub-confluent monolayer. Cells were rinsed with PBS and fixed with 100% ethanol for 20 min at room temperature. Subsequently, the cells were saturated for 45 min with PBS containing 2% BSA and were incubated for 1 h at 37°C with 10-20 μg/ml rabbit polyclonal antibodies against either Cx26 or Cx43 (Zymed Laboratories, San Fracisco CA), diluted in PBS containing 2% BSA. After extensive washing with PBS, cells were incubated for 1 h with a 1:40 dilution of a FITC-conjugated porcine anti-rabbit IgG (Dako, Carpinteria CA). The cells on coverslips were washed three times with PBS, mounted on glass slides, and viewed under an Olympus fluorescence microscope.
SVEC4-10 cells transiently expressing HSVtk were mixed in a 1:9 ratio with non-transduced tumor cells as indicated and were grown to a sub-confluent layer. The co-cultures were treated with 20 μM GCV. In subsequent experimental settings, the long-term inhibitor of GJIC, 18-α-glycyrrhetinic acid (AGA; Sigma), was added at 70 μM to the medium during the treatment with GCV. Media containing GCV, AGA or both was renewed every 2 days and the viable cells were counted after 5 days by the Trypan blue-exclusion methodology.
Targeted RGD-4C AAVP particles as well as non-targeted controls were amplified, isolated and purified from the culture supernatant of host bacteria (E. coli MC1061) as we previously described (20, 21, 28, 31). Next, vector particles in suspension were sterile-filtered through 0.45-μm filters, then titrated by infection of host bacteria for colony counting on Luria–Bertani (LB) agar plates under a double antibiotic selection and expressed as bacterial transducing units (TU).
Tumor-bearing mice were established and tumor volumes were calculated as described (20, 21, 29). Mice were anesthetized by gas (2% isoflurane and 98% oxygen) inhalation. Tumor cells were released by exposure to trypsin, counted, centrifuged, and resuspended in serum-free medium. A total of 5 × 104 cells from the EF43-FGF4 mouse mammary tumor were implanted subcutaneously into 6 week-old female BALB/c immunocompetent mice. When tumors reached a volume of ~100 mm3, tumor-bearing mice received a single intravenous dose of RGD-4C AAVP-HSVtk (5 × 1010 TU), or controls. Treatment with GCV (80 mg/kg per day, intraperitoneal) was initiated later in cohorts of size-matched, tumor-bearing mice as indicated. Tumor growth was monitored daily and measured by caliper twice weekly. Each experimental cohort contained at least 15 tumor-bearing mice divided into 3 groups of 5 mice each. Simple hypothesis test (Mann-Whitney test statistic) was applied to assess whether differences among groups were significant. Statistical significance level was set to α= 0.05.
The Institutional Animal Care and Use Committee approved all experimentation described here. Mice were anesthetized by 2% isoflurane and 98% oxygen inhalation. Cultured cells were detached with a solution of trypsin-EDTA, counted, centrifuged, and resuspended in serum-free medium. The EF43-FGF4 tumor cells (5×104) were mixed with HSVtk-transduced SVEC4-10 (5×104) or parental SVEC4-10 cells (5×104), then the resulting cell mixtures (1×105 cells at a 1:1 ratio) were injected subcutaneously into the back of 6-week-old female athymic nu/nu (nude) mice. Treatment of mice with GCV started 6 days after cell implantation into mice and was administered daily by intraperitoneal injection of 80 mg/kg. Tumor growth was monitored three times a week by caliper measurement of 2 diameters, and expressed as mean tumoral volume ± SD. In each assay, the number of mice per group (n) was 10.
To understand the lack of transduction of EF43-FGF4 breast cancer cells by the targeted RGD-4C AAVP vector, we assessed the expression of the receptors of RGD-4C ligand, αv integrins, in EF43-FGF4 cells. We carried out fluorescence-activated cell sorting (FACS) analysis in vitro. The data revealed that EF43-FGF4-derived tumor cells barely express the αv integrins on their surface (Fig. 1A). Kaposi's sarcoma cells (KS1767), which served as a positive control (20, 23, 24), showed strong expression of αv integrins. The corresponding negative control, in which species-matched IgG isotype control antibodies were used, lacked αv integrin expression. These data are consistent with our findings of non-transduction of EF43-FGF4 cells by the RGD-4C AAVP vector (data not shown).
For assessment of the therapeutic efficacy of the RGD-4C AAVP-HSVtk vector on the growth of EF43-FGF4 tumors in vivo, mice with established EF43-FGF4 tumors (~100 mm3) received 5×1010 transducing units (TU) of targeted RGD-4C AAVP-HSVtk intravenously. Treatment with GCV (80 mg/kg/day, intraperitoneally) was initiated 2 days later. There was marked growth suppression of established tumors in the presence of targeted RGD-4C-displaying vector but not with non-targeted control vectors (Fig. 1B); moreover, tumor growth was not affected in several negative experimental control groups (Fig. 1B). For statistical analyses, we compared the median tumor growth of the group treated with RGD-4C AAVP-HSVtk against all other seven control groups (Fig. 1B). For all comparisons we applied the non-parametric Mann-Whitney test statistic. All P values were equal to 0.0079 (significant at significance level α = 0.05).
These results establish that targeted suicide gene therapy to the tumor tissue after systemic administration of the RGD-4C AAVP vector results in efficient anti-tumor therapy, despite the lack of direct tumor cell transduction.
First, we determined that the EF43-FGF4 cells can be killed by a bystander effect in vitro. EF43-FGF4 cells transiently expressing HSVtk (100%) were mixed in a 1:9 ratio with parental non-transduced (i.e., HSVtk-negative) EF43-FGF4 cells and grown in co-culture to sub-confluent monolayers. Addition of 20 μM GCV resulted in cell death in over 98% of the co-culture, as confirmed by Trypan blue exclusion (data not shown). This suggests a potent bystander effect among EF43-FGF4 cells. Next, we investigated a potential heterotypic bystander effect between the EF43-FGF4 tumor cells and HSVtk-transduced SVEC4-10 mouse endothelial cells from an SV40-transformed murine small blood vessels. We chose these cells after verifying that they display a strong bystander effect in homoculture after transduction with HSVtk and subsequent GCV treatment (Fig. 2A). SVEC4-10 cells transiently expressing HSVtk were grown in a mixed heteroculture with non-transduced (i.e., HSVtk-negative) EF43-FGF4 cells (1:9 ratio). After 5 days of GCV treatment, over 90% of the cells in the heteroculture were killed indicating a strong bystander effect between HSVtk-transduced endothelial cells and the non-transduced tumor cells (Fig. 2B). To show that this result was not a unique phenomenon between endothelial cells and a particular tumor cell line, we reproduced the effect with malignant glioma 9L cells (Fig. 2C). The cell line MDA-MB435 served as a negative control, as it does not display bystander killing in homoculture (data not shown). Consistently, no bystander killing was observed between MDA-MB435 cells and HSVtk-transduced endothelial cells (Fig. 2D).
Next, we attempted to identify the mechanism of the observed heterotypic bystander effect between HSVtk-expressing endothelial cells and tumor cells. Gap junctional intercellular communication (GJIC) plays a central role in mediating bystander effects (32). Therefore, we investigated GJIC function in our heterotypic co-culture system and analyzed the effect of a selective GJIC inhibitor, 18α-glycyrrhetinic acid (AGA), on the heterotypic bystander killing that was observed in vitro. In heterotypic co-cultures of 1:9 HSVtk-transduced SVEC4-10 endothelial cells and non-transduced EF43-FGF4 cells, or 9L cells, respectively, the addition of 70 μM AGA substantially inhibited bystander killing upon GCV treatment (Fig. 3). For statistical analyses, we compared in each co-culture the mean of cell survival between control and GCV treatment as well as between GCV and AGA treatments (Fig. 3). We applied the t-test statistic and for all pair-wise comparisons P < 2.2 × 10−16 (significant at significance level α = 0.05).
This result indicates that the heterotypic bystander effect is related to GJIC. It therefore became relevant to analyze the expression of Cx43 and Cx26 in the cell types involved in the bystander effect. Connexins are proteins composing the channels of the GJIC through which toxic phosphorylated GCV and/or other toxic intracellular metabolites are exchanged between one cell and another (33, 34). Immunofluorescence staining showed strong expression of both Cx43 and Cx26 in all cell types that exhibited bystander effect (Fig. 4). Consistently with our hypothesis, MDA-MB435 cells, which are not susceptible to the bystander effect, did not express these connexins (Fig. 4). We conclude that gap junctions mediate the heterotypic bystander effect between tumor and endothelial cells and that such cell junctions contain the connexins Cx43 and Cx26.
Finally, we determined whether the bystander killing that was induced in EF43-FGF4 tumor cells by the HSVtk-transduced SVEC4-10 endothelial cells in vitro could also be observed in vivo. Therefore, HSVtk-expressing SVEC4-10 or untransduced SVEC4-10 cells, respectively, were mixed at a 1:1 ratio with EF43-FGF4, or SVEC4-10 cells and then implanted into nude mice. Tumors were formed only from cell mixtures containing EF43-FGF4 cells, whereas the SVEC4-10 cells alone were non-tumorigenic. Systemic treatment with GCV started at day #6 after implantation, and the drug was administered daily until day #20. Mice injected with the control cell mixture of parental (HSVtk-negative) SVEC4-10 and EF43-FGF4 cells developed large tumors with rapid invasive growth (Fig. 5), comparable to that of homotypic parental EF43-FGF4 tumors. In contrast, almost total inhibition of tumor growth was achieved in mice implanted with SVEC4-10-HSVtk-positive and EF43-FGF4 cell mixtures (Fig. 5). These results establish that the heterotypic bystander effect between endothelial and tumor cells can also be elicited in vivo.
Tumor vascular endothelium has certain properties that render it an attractive target for cancer gene therapy such as accessibility to circulating vectors, expression of endothelial surface receptors distinct from those of normal quiescent vasculature, and a potential amplifying effect caused by hypoxia. Ligand-directed AAVP can mediate targeted HSVtk suicide gene transfer to tumor vascular endothelium in several experimental models (20, 21). However, the precise cytotoxic mechanism of such profound anti-tumor effects after vascular-targeted suicide gene transfer has not as yet been entirely understood. Here we show that vascular targeted HSVtk suicide gene delivery results in efficient cell killing mediated by a heterotypic bystander effect between endothelial and parenchymal tumor cells in vitro and in vivo.
The bystander effect has initially been described as a phenomenon in homotypic cultures, in which neighbouring (“bystander”) HSVtk-non-expressing tumor cells were also killed by GCV (1, 2, 35, 36). Later, similar bystander effects were reported in other suicide gene systems (37), and it was observed after ionizing radiation as well (38). These observations suggest that not only activated cytotoxic drugs, but also other toxic metabolites, can be transferred from treated to adjacent untreated cells. While well documented in vitro, the bystander effect after suicide gene transfer has been less studied in animal models (1, 35, 39), and only in experimental systems in which the parenchymal tumor cells were the primary target of the gene transfer. Clearly, there is room to improve the knowledge about this particular cell killing mechanism in an in vivo setting and perhaps to develop new translational applications.
We show that gap junctions mediate the heterotypic bystander effect in our system, as it can be blocked by AGA, a potent inhibitor of gap junctional intracellular communication. The intercellular junctions formed by both endothelial cells and tumor parenchymal cells contain Cx43 and Cx26; overexpression of which in gap junctions have been shown to potentiate the bystander effect (33, 34, 40). Consequently, gene transfer-mediated forced expression of these connexins in cells with low levels of gap junctions can result in potent induction of a bystander effect in cells lacking expression of the suicide gene (41). Therefore, the expression of these connexins selectively in the cell types displaying bystander effects in this study suggests that these proteins are involved in the heterotypic bystander killing described here. It is generally assumed that the level of the bystander effect is determined by the characteristics of the non-HSVtk-transduced cell population (40, 41). Consistent with this assumption, we did not observe bystander killing in co-cultures of connexin-expressing HSVtk-transduced endothelial cells and non-connexin-expressing, non-transduced MDA-MB435 cells.
The heterotypic bystander effect between endothelial and epithelial tumor cells can also be induced in vivo. Thus, such phenomenon likely accounts for the extend of tumor cell killing observed in various tumor cell models after endothelial cell-directed suicide gene transfer by targeted AAVP vector and potentially other vascular-targeted gene therapy vectors. Often, a regular tumor graft model in which the endothelium is destroyed by vascular-targeted suicide gene therapy followed by secondary tumor eradication, does not allow a rigorous dissection of underlying cytotoxic mechanisms. The model we have used in this study potentially circumvents this limitation. Co-administration of tumor cells and endothelial cells results in the formation of chimeric tumors, in which a pre-determined fraction of the cells is endothelial, but it does not contribute to the circulation that is host-dependent. Notably, killing of the graft endothelial cell population by suicide gene transfer and subsequent treatment with GCV can most likely be explained by a heterotypic bystander effect rather than indirect tumor killing due to destruction of the vasculature and subsequent hypoxia. Such heterotypic bystander effects have been considered but not proven in previous models (36), and our study supplies for the first time systematic evidence to show that this hypothesis is valid. Nevertheless, additional mechanisms mediating tumor cell killing after suicide gene transfer in vascular endothelial cells in vivo cannot be excluded. Such additional conceivable mechanisms include bystander cell phagocytosis of apoptotic factors released into the extracellular space by dying cells (35, 42) or host immune responses following the HSVtk plus GCV treatment (39, 43-45). As such, one must speculate that many of these putative mechanisms are non-mutually exclusive, may be context-dependent, and may also occur between tumor cells and other non-vascular tumor stromal cells.
In summary, our demonstration of a heterotypic bystander effect in vivo may have implications for cancer gene therapy. Systemic, effectively targeted suicide gene delivery to non-parenchymal cells within a tumor may yield significant tumor responses in preclinical systems, even if the targeted cell population constitutes only the genetically non-malignant fraction of the tumor. Together, these data add another potential cytotoxic mechanism to suicide gene therapy based on targeted AAVP, and support the promising role of non-malignant tumor vascular and/or stromal cells as candidate therapeutic targets.
We thank Dr. Mohamed Trebak for comments on the manuscript, Dr. Frank C. Marini for discussions and Catherine A. Moya for technical assistance.
This work was supported by grants CA90270 and CA8297601 (to R.P.), CA90270, CA70907, and CA9081001 (to W.A.), by grant TR448/5-3 (to M.T.) from the Deutsche Forschungsgemeinschaft, and by awards from the Gillson-Longenbaugh Foundation (to R.P. and W.A.), the W.M. Keck Foundation (to R.P. and W.A.), the Susan G. Komen Breast Cancer Foundation (to M.T.) and by a grant G0701159 of the UK Medical Research Council (to C.S. and A.H.)
Disclosure of Potential Conflicts of Interest:
There is no potential conflict of interest relevant to this article