SiRNA-mediated Inhibition of Vascular Endothelial Growth Factor Severely Limits Tumor Resistance to Antiangiogenic Thrombospondin-1 and Slows Tumor Vascularization and Growth¹

The observation that tumor growth is highly dependent on the ability of tumors to induce their own vascularization has led numerous laboratories to isolate or develop angiogenesis inhibitors such as TSP1⁴ (1). This antivascular strategy that targets the normal, genetically stable endothelial cells of the host rather than the genetically unstable tumor cell population was shown to be very efficient at reducing tumor growth and was not expected to trigger tumor resistance (2). However, recently we (3) and others (4) have demonstrated that changes in the tumor cells themselves, particularly sustained high-level secretion of the angiogenic stimulator VEGF, can enable tumors to bypass antiangiogenic treatments.

It has recently been shown that the introduction in a mammalian cell of double-stranded oligoribonucleotides, also called siRNA, triggers the degradation of the endogenous mRNA to which the siRNA hybridizes (5). This mechanism is highly sequence specific and allows to turn off the expression of a target protein (6, 7). Many studies demonstrated the high efficiency and versatility of RNA interference in cell cultures. Some authors developed vectors or viruses to produce siRNA in cells (8). The in vivo regulation of a gene by RNA interference has been obtained either using these vectors or viruses (9) or using the so-called hyperpressure technique (10, 11), which drives siRNA mainly in the liver and would not be possible to use in humans. In this work, we demonstrate that low doses of siRNA administrated by a systemic route penetrate into tumors and control the expression of target genes to produce phenotypic effects.

The aim of the present work was to determine whether blocking the ability of tumor cells to secrete high levels of VEGF by the in vivo administration of siRNA could diminish or prevent the triggering of resistance to the antitumor effects of TSP1.

The rat fibrosarcoma cJ4 cells (12) were grown as described previously. A bidirectional TSP1-luciferase-inducible expression vector was introduced in cJ4 cells to generate JT8 cells as described previously (3). Cells were grown in the presence of 100 ng/ml dox, a tetracycline analogue to repress TSP1 expression.

The siRNA (sense and antisense strands) were purchased from MWG Biotech (Ebersberg, Germany). The sense strands sequences were the following: VEGF, 5′-AUGUGAAUGCAGACCAAAGAA-TT; CONT, 5′-GAUAGCAAUGACGAAUGCGUA-TT; and LUC, 5′-AACGUACGCGGAAUACUUCGA-TT. In vitro transfections were performed using the Transit-TKO polymer/lipid from Mirus (Madison, WI) as recommended. For 6 × 10⁶ cells in 10 ml of medium, 2 μg of siRNA were used. Cells were washed 24 h after transfection.

cJ4 or JT8 cells were injected s.c. in PBS (10⁶ cells/site) into the hind quarters of four to six female Swiss nu/nu mice, 4–6 weeks old (Iffa Credo, L’Arbresle, France) for each tested condition. Each experiment was repeated at least twice. When stated, dox (100 μg/ml) was added to the drinking supply of the animals to repress TSP1 and luciferase expressions. The drinking supply was changed three times a week. Tumor volume was calculated as v = L × l² × 0.52, where L and l represent the larger and the smaller tumor diameter measured daily. For in vivo injections, each animal was injected daily with 50 (i.p., i.v., or s.c. injections) or 10 μl (intratumoral injections) of PBS containing 3 μg of siRNA (125 μg/kg/day). The care of the animals was provided in the animal quarters of the Institut André Lwoff in Villejuif according to the institutional guidelines.

TSP1, VEGF, and CD31 detection and scoring of blood vessels density in tumors were performed as described previously (3).

Tumors were homogenized with a polytron homogenizer in cell culture lysis reagent (Promega). Protein concentration was measured using BSA as standard with the Bio-Rad DC protein assay. Luciferase activity was quantified in a luminometer (Analytical Luminescence Laboratories) using 1 mm luciferine as substrate.

VEGF was quantified in cell supernatants or in tumor homogenates using an ELISA kit for mouse VEGF from R&D. Cells were transfected with VEGF- or control-siRNA. Twenty-four h later, cells were replated at the same density and conditioned media collected on day 4 after transfection. Cells were replated at equal density on day 4 and media collected on day 6 and finally replated on day 11 and media collected on day 13. Values are expressed as percentage of the VEGF content in the CONT-siRNA-transfected cells medium collected on the same day. VEGF in tumor homogenates is expressed as pg/mg of total protein.

siRNA matching a 21-nt sequence conserved between the human, rat, and mice VEGF A mRNA was synthesized (VEGFsiRNA). As controls, we used either a sequence presenting no significant homology with mRNA databases (CONTsiRNA) or a siRNA against luciferase mRNA (LUCsiRNA).

Transfection of VEGFsiRNA in cJ4 rat fibrosarcoma cell line (12) induced a marked reduction in VEGF synthesis and secretion (Fig. 1,a, bottom panel, and b) as compared with cells transfected with the CONTsiRNA (Fig. 1,a, left panel). The secreted VEGF level was still reduced by 50% 13 days after transfection. Untransfected, CONTsiRNA-, or VEGFsiRNA-transfected cells were engrafted s.c. to nude mice and tumor growth monitored (Fig. 1,c). On day 12, animals were sacrificed and tumors collected. Immunodetection of VEGF in the tumors showed a marked reduction in the VEGF expression of tumors growing from VEGFsiRNA-transfected cells (Fig. 1,d, right panel) as compared with controls. This reduction was accompanied by a 67% reduction in tumor volume. These data indicated that the reduction in VEGF synthesis obtained by the siRNA transfection in vitro resulted in the expected biological effects on VEGF production and tumor growth in vivo. The tumor growth was unaffected by the transfection of control siRNA. The logarithmic regression analysis of the tumor growth curves (Fig. 1 c, inset) shows that the onset of tumors that express only residual VEGF levels were delayed as compared with controls but eventually grew with similar growth rates.

To monitor the delivery of siRNA to tumors by systemic administration, we first implanted in nude mice cJ4-derived cells constitutively expressing luciferase. When tumors reached ∼200 mm³, mice were separated into groups of three, and each mouse received a single injection of LUCsiRNA or CONTsiRNA. siRNA was administered either via the tail vein i.v., i.p., or s.c. (3 μg in 50 μl of saline in each case) or directly in the tumor (3 μg in 20 μl of saline). Three days later, animals were sacrificed, tumors collected, homogenized, and luciferase activity and protein content were determined. The three systemic administration procedures induced ∼50% inhibition of the luciferase expression (Fig. 2), whereas the direct intratumoral injection was ineffective.

Recently, using the same parental fibrosarcoma line, we established JT8 cells where the TSP1 sequence is expressed from a bidirectional vector in which a common tet operator controls TSP1 and luciferase. So, in vitro as in vivo, in the absence of dox (a tetracycline analogue), TSP1 and luciferase are expressed, whereas when dox is added either to the medium of the cultured cells or in the drinking supply of the animals if the cells are growing in vivo, both genes are repressed. We assayed the effect of VEGF-siRNA delivered daily (125 μg/kg/day) by i.p. administration on the growth of JT8 cells grafted s.c. to nude mice receiving dox that insured TSP1 and luciferase were repressed. By day 16, the volume of tumors growing in such treated animals (group B) was reduced by 66% as compared with the controls (group A), which had been treated with LUCsiRNA (Fig. 3,a). A 70% reduction in VEGF expression was measured by ELISA in homogenates of tumors from the VEGFsiRNA-treated group (Fig. 3,b). To control that this reduction was not resulting from a nonspecific siRNA effect, the luciferase activity was measured in the same tumor homogenates. As expected in the presence of dox, this activity was low but measurable with a luminometric assay. We measured a 52% reduction in luciferase activity in LUCsiRNA- as compared with VEGFsiRNA-treated tumors (Fig. 3 b). Only a few number of studies describe the use of RNA interference in vivo by retroviral (9) or hydrodynamical (10, 11) procedures. We demonstrate here the efficient delivery into tumors of siRNA by systemic i.p. administration at low doses (125 μg/kg/day) in saline. This treatment resulted in a robust inhibition of the endogenous VEGF expression and produced a marked tumor growth inhibition. No side effects were detected on the living animals or on their organs observed at the end of the experiment.

The efficiency of i.p. administration of VEGFsiRNA on tumor growth inhibition was compared with the effect of TSP1 expression by the tumor cells. A third group of mice (group C) received no dox supply, ensuring the expression of TSP1 and luciferase, and was injected with LUCsiRNA. The growth curves of tumors from groups B and C were very similar (Fig. 3,a). The VEGF dosage indicated that tumors from group C grew from cells overexpressing VEGF 2.7-fold as compared with tumor cells growing in the absence of TSP1 (Fig. 3,b). The tumors in the control group reached a volume of 100 mm³ 9.7 ± 0.9 days after the injection of the cells and 3 days later when TSP1 was expressed (12.7 ± 0.7 days) or VEGF inhibited (12.7 ± 0.5 days) in the tumors. The mean of the tumor doubling times in untreated animals (40.4 ± 2.5 h) was not significantly modified by TSP1 (41.0 ± 4.8 h) or VEGFsiRNA treatments (44.4 ± 1.7 h; Fig. 3 a, inset). These data demonstrate that as previously observed for TSP1 expression, VEGF inhibition by siRNA delays the onset of the tumors but has no effects on the growth rates of the tumors.

To test the efficiency of i.p. administration of VEGFsiRNA on VEGF expression that is triggered by the in vivo development of resistance to TSP1 expression, finally, in a fourth group (D), the two treatments were combined: JT8 tumors were grown in animals receiving no dox and injected with VEGFsiRNA. The inhibition of the tumor growth observed in this group was not better than that obtained with VEGFsiRNA alone or TSP1 alone (Fig. 3,a). We suspected that TSP1 might be reducing the accessibility of the siRNA to the tumors by limiting tumor vascularization. Luciferase expression was then measured in the tumors from groups C and D. As expected, in the absence of dox treatment, the luciferase activity, which parallels TSP1 expression, was increased 6.8-fold in tumors from group D as compared with group A. Although the animals in group C were injected with LUCsiRNA, no decrease in the luciferase activity was observed in tumors from this group, and only a 23% reduction in VEGF expression was triggered by VEGFsiRNA treatment in group D as compared with group C (Fig. 3 b). This strongly suggested that the siRNA was not penetrating the tumor properly. This result points out the difficulties likely to be encountered whenever one adds a systemic treatment to an antiangiogenic therapy and offers an explanation for why tumors injected daily with VEGFsiRNA are eventually able to grow.

To circumvent this difficulty of delivering systemic siRNA to tumors in the presence of TSP1, JT8 cells were transfected with VEGFsiRNA or LUCsiRNA and injected to nude mice receiving no dox treatment, thus expressing TSP1. On day 16, the mean tumor volume of the control group was 799 ± 270 mm³ (Fig. 4), very close to the values observed in the previous experiment (Fig. 3,a, group C). This volume was reduced by 86% when the cells were unable to produce VEGF efficiently because of the presence of VEGFsiRNA (mean tumor volume, 111 ± 70 mm³). In these barely palpable tumors, the VEGFsiRNA induced an 82% reduction in VEGF expression (9.8 ± 2.7 versus 56.4 ± 23.9 pg/mg protein). Cells that grew in vivo despite these treatments could be untransfected cells, cells re-expressing VEGF once the siRNA effect is diluted by exponential cell divisions or cells triggering angiogenesis by elaborating other angiogenic factors such as basic fibroblast growth factor. The tumors in this group reached 100 mm³ 15.2 ± 1.2 days after injection of the tumor cells, indicating additive effects of TSP1 and VEGFsiRNA to delay the onset of tumors. Remarkably, the logarithmic regression analysis of the growth curves show that VEGFsiRNA also reduced the growth rate of the tumors (Fig. 4, inset). The doubling time of tumors treated with TSP1 alone (41.9 ± 2.7 h) was increased by 65% when VEGFsiRNA were used in combination with TSP1 (69.3 ± 10.3 h). This combination thus not only delays by 6 days the onset of the tumors but also significantly slows down their growth, a parameter that was not affected by the administration of one single treatment.

Numerous antiangiogenic agents are currently in clinical trials. Monotherapies in these assays have been disappointing, and several companies have abandoned drugs targeting VEGF or its receptors (13). Resistance, similar to that which has been documented in mice, could be responsible for the failures of these promising agents to consistently curtail human tumor growth. If so, combinational therapies appear to be needed (14, 15). The joining of the antiangiogenic protein endostatin with an antisense strategy against the epidermal growth factor receptor (expressed on a fraction of human tumor cells) has been shown to produce synergistic inhibitory effects on tumor growth (16). In this study, we show that the combination of two different antiangiogenic agents can produce synergistic and not only additive effects to significantly minimize tumor resistance and tumor growth rate. Our finding that it is possible to use siRNA in vivo in such studies should facilitate additional efforts to define more efficient sets of treatments. Moreover, siRNA may be used to suppress expression of point mutated genes frequently appearing during the natural history of cancers in humans (17). It is not clear today if the uptake of siRNA in a tumor is facilitated as compared with normal tissues, and biodisponibility studies will have to be performed to address this question. Because of their transient effects in vivo and their capacity to penetrate a tumor when injected via a systemic route, it may eventually be possible to use siRNAs sequentially to target one after the other different key proteins in the development of cancer and thereby keep tumors under control.

We thank Noël Bouck for critical review of this work and very useful comments and Lauriane Fritsch and Cécile Goujet-Zalc for some interesting suggestions.