Polyethylenimine/small interfering RNA-mediated knockdown of vascular endothelial growth factor in vivo exerts anti-tumor effects synergistically with Bevacizumab

Abstract

Background

RNA interference is a powerful method for the knockdown of pathologically relevant genes. The in vivo delivery of siRNAs, preferably through systemic, nonviral administration, poses the major challenge in the therapeutic application of RNAi. Small interfering RNA (siRNA) complexation with polyethylenimines (PEI) may represent a promising strategy for siRNA-based therapies and, recently, the novel branched PEI F25-LMW has been introduced in vitro. Vascular endothelial growth factor (VEGF) is frequently overexpressed in tumors and promotes tumor growth, angiogenesis and metastasis and thus represents an attractive target gene in tumor therapy.

Methods

In subcutaneous tumor xenograft mouse models, we established the therapeutic efficacy and safety of PEI F25-LMW/siRNA-mediated knockdown of VEGF. In biodistribution and siRNA quantification studies, we optimized administration strategies and, employing chemically modified siRNAs, compared the anti-tumorigenic efficacies of: (i) PEI/siRNA-mediated VEGF targeting; (ii) treatment with the monoclonal anti-VEGF antibody Bevacizumab (Avastin®); and (iii) a combination of both.

Results

Efficient siRNA delivery is observed upon systemic administration, with the biodistribution being dependent on the mode of injection. Toxicity studies reveal no hepatotoxicity, proinflammatory cytokine induction or other side-effects of PEI F25-LMW/siRNA complexes or polyethylenimine, and tumor analyses show efficient VEGF knockdown upon siRNA delivery, leading to reduced tumor cell proliferation and angiogenesis. The determination of anti-tumor effects reveals that, in pancreas carcinoma xenografts, single treatment with PEI/siRNA complexes or Bevacizumab is already highly efficacious, whereas, in prostate carcinoma, synergistic effects of both treatments are observed.

Conclusions

PEI F25-LMW/siRNA complexes, which can be stored frozen as opposed to many other carriers, represent an efficient, safe and promising avenue in anti-tumor therapy, and PEI/siRNA-mediated, therapeutic VEGF knockdown exerts anti-tumor effects. Copyright © 2010 John Wiley & Sons, Ltd.

Introduction

RNA interference (RNAi) allows the knockdown of any particular gene of interest, offering great potential as a novel therapeutic strategy. It is mediated through approximately 21–23 nt, double-stranded ‘small interfering RNAs’ (siRNAs), which trigger the sequence-specific cleavage of mRNA molecules leading to their subsequent degradation 1. These siRNAs are generated intracellularly through the cleavage of longer double-stranded RNAs 2, 3 or can be directly introduced into the cell as chemically synthesized siRNA molecules 1.

Vascular endothelial growth factor (VEGF) is one of the principal regulators of tumor growth, tumor angiogenesis and metastasis, and is present at significant levels in tumor cells from various origin. Examples include neoplastic, but not benign hyperplastic or normal prostate cells 4, 5, as well as pancreatic tumor cells 6. Inhibition of VEGF or its receptor (VEGFR) inhibits tumor growth and metastasis in animal tumor models 7, 8. Consequently, several approaches have been used to implement the blockade of VEGF or VEGF receptor signaling in cancer treatment, including humanized antibodies to VEGF 9, as well as gene targeting approaches through RNA interference (RNAi). Besides the blockage of tumor angiogenesis upon VEGF inhibition, autocrine/paracrine VEGF/VEGFR signaling, which results in the stimulation of tumor cell proliferation, has also been shown 10–13. Bevacizumab (Avastin®, Roche, Basel, Switzerland), a humanized monoclonal anti-VEGF antibody, is the first VEGF inhibitor approved by the FDA for systemic use in cancer 14 and shows activity in various cancers, including pancreas and prostate carcinoma 15.

Upon RNAi-mediated VEGF down-regulation through in vitro transfection, anti-proliferative and/or anti-tumoral effects in various tumor cell lines including prostate and pancreatic carcinoma cells were oberved 16–18. In vivo, however, synthetic siRNAs display poor stability and poor penetration into the cells and, thus, the main issue is the delivery of therapeutically active siRNAs into the target tissue/target cells 19. To circumvent these problems, the approach of direct intratumoral injection has been employed for (formulated) siRNA molecules in subcutaneous (s.c.) tumor models 20–24 with the local administration, however, probably being of only limited therapeutic relevance in cancer therapy. Alternatively, the viral or nonviral delivery of DNA-based siRNA constructs for RNAi-mediated VEGF down-regulation showed anti-tumorigenic effects in various xenograft models 25–28. However, limited loading capacities, problems in large-scale production and, most importantly, safety risks as a result of their inflammatory and immunogenic effects and their oncogenic potential pose severe limitations to the applicability of viruses. Issues regarding safety and efficacy are also critical with regard to the use of plasmid-based siRNA constructs in vivo. Thus, the direct, systemic, nonviral administration of siRNA molecules allowing their therapeutic use is most desirable.

Previously, we and others have introduced polyethylenimine (PEI) complexation of siRNAs as an efficient tool for in vivo siRNA delivery 19. PEIs are synthetic linear or branched polymers available in a wide range of molecular weights 29, 30 and, because of their high cationic charge density at physiological pH, they are able to form noncovalent complexes with siRNAs 31, 32. This complexation leads to siRNA protection against degradation, efficient cellular uptake through endocytosis and subsequent intracellular release from endosomes based on the so-called ‘proton-sponge effect’ 33, 34. Using the commercially available, linear jetPEI, anti-tumorigenic effects were demonstrated upon targeting of the HER-2 receptor in s.c. ovarian carcinoma xenografts 31 or the growth factor pleiotrophin in s.c. or orthotopic glioblastoma xenografts 32. Modified PEI coupled to polyethylenglycol in combination with a ligand for tissue-specific targeting (RGD peptide for the recognition of tumor vasculature) was used for VEGFR knockdown 35, and polyelectrolyte complex micelles have been employed for VEGF targeting in a s.c. tumor model 36.

Only certain PEIs are capable of successfully delivering siRNAs and, more recently, Werth et al.37 described the preparation of PEI F25-LMW, a low molecular weight 4–10 kDa branched PEI. The in vitro data demonstrate low toxicity, high cellular uptake efficacy and efficient protection/intracellular release of siRNA molecules 37, as well as favorable features regarding the preparation, handling and long-term storage of the complexes. In particular, without the addition of any lyoprotectant, PEI F25-LMW/siRNA complexes can be stored frozen for several months with full retention of their in vitro bioactivity 38.

In the present study, we explored the in vivo efficacy and safety of PEI F25-LMW-mediated gene targeting of VEGF in mouse tumor xenograft models and present optimized administration strategies by performing biodistribution studies. Using novel, chemically modified siRNAs, we analysed the anti-tumorigenic effects of PEI/VEGF siRNA or Bevacizumab treatment alone, as well as the synergistic effects of a combination of both, and establish PEI F25-LMW/VEGF siRNA-mediated VEGF targeting as a novel strategy in tumor therapy.

Materials and methods

siRNAs, tissue culture and animals

VEGF targeting siRNAs were chemically synthesized with modifications, as described previously 39. The target specific siRNAs were 4332 with the passenger strand 5′-cAuAGGAGAGAuGAGcuucdT*dT-3′ and the guide strand sequence 5′-GAAGCUCAUCUCUCCUAUGdT*dT-3′, and 4371 with the passenger strand sequence 5′-GuGAAuGcAGACcAAAGAAA*G-3′ and the guide strand sequence 5′-CUUUCUUuGGUCuGcAUUcAC*A*U-3′. These sequences were selected based on a screening of a larger set of siRNAs (data not shown). The negative control siRNA 5246 was directed against the green fluorescence protein (GFP) and consisted of the passenger strand sequence 5′-CcAcAuGAAGcAGcACGACu*U-3′ and the guide strand sequence 5′-AAGUCGuGCuGCuUcAUGuGG*u*C-3′. In the oligoribonucleotide sequences, upper case letters represent ribonucleotides, lower case letters represent 2′-O-methyl-nucleotides, upper case letters printed in italics represent 2′-deoxy-2′-fluoro-nucleotides, the asterisk (*) represents the phosphorothioate linkages and dT represents deoxythymidine nucleotides. For comparison, the chemically unmodified VEGF-specific siRNA (passenger strand 5′-GGAGGAGGGCAGAAUCAUCdTdT-3′; guide strand 5′-GAUGAUUCUGCCCUCCUCCdTdT-3′; Ambion/Applied Biosystems, Darmstadt, Germany) was used. PANC-1 pancreas carcinoma and PC-3 prostate carcinoma cells were obtained from the American type culture collection (ATCC) and cultivated in a humidified incubator under standard conditions (37 °C, 5% CO₂) in IMDM (PAA, Cölbe, Germany) supplemented with 10% fetal calf serum (FCS). Athymic nude mice (nu/nu) were purchased from Harlan Winkelmann (Borchen, Germany) and kept at 23 °C in a humidified athmosphere with food and water available ad libidum. Animal studies were approved by the Regierungspräsidium Giessen, Germany.

PEI complexation

PEI F25-LMW/siRNA complexes were prepared essentially as described previously 37. Briefly, 100 pmol (1.3 µg) of siRNA was dissolved in 80 µl of 10 mM HEPES/150 mM NaCl, pH 7.4, and incubated for 10 min. Twenty-two microliters of PEI F25-LMW (0.6 µg/µl) 37 was dissolved in 80 µl of the same buffer and, after 10 min, pipetted into the siRNA solution. This resulted in an N/P ratio = 33, which was determined as optimal for siRNAs in pilot experiments (not shown). For in vivo experiments, the mixture was aliquotted and stored frozen. Prior to use, complexes were thawed and incubated for 1 h at room temperature.

In vitro transfections and growth assays

PC-3 or PANC-1 cells were seeded at 1 × 10⁵ cells/well (six-well plate), 5 × 10⁴ cells/well (24-well plate, 3 × 10³ cells/well [96-well plate, enzyme-linked immunosorbent assay (ELISA)] or 1 × 10³ cells/well (96-well plate, proliferation assay), 24 h prior to transfection. Transfections were performed by the addition of the specific or nonspecific PEI F25-LMW/siRNA complex (100, 60 or 10 pmol of siRNA/well of a six-well, 24-well or 96-well plate, respectively). For RNA quantification, cells from sixwell-plates were harvested 72–96 h after transfection and, for VEGF protein quantification, medium was changed to IMDM/2% FCS and harvested after additional 3–5 days. Studies of anchorage-dependent proliferation were carried out essentially as described previously 40 in the presence of IMDM/2% FCS. At the time points indicated, the numbers of viable cells in eight wells were determined using a colorimetric assay according to the manufacturer's protocol (Cell Proliferation Reagent WST-1; Roche Molecular Biochemicals, Mannheim, Germany). Anchorage-independent proliferation was studied in soft agar assays essentially, as described previously 41. Cells were transfected in six wells with PEI-complexed specific or nonspecific siRNAs as described above and trypsinized and counted 24 h after transfection. Twenty thousand cells in 0.35% agar (Bacto Agar, Becton Dickinson, Franklin Lakes, NJ, USA) were layered on top of 1 ml of a solidified 0.6% agar layer in a 35-mm dish. Growth media with 2% FCS were included in both layers. After 2–3 weeks of incubation, colonies more than 50 µm in diameter were counted by at least two independent blinded investigators.

RNA preparation and quantitative reverse transcriptase-PCR (RT-PCR)

Total RNA from tumor cells or homogenized tissues was isolated using the Tri reagent (PEQLAB, Erlangen, Germany) according to the manufacturer's instructions. For tissue homogenization prior to RNA preparation, tissues were mixed with 200 µl of Tri reagent and homogenized. RT was performed using the RevertAid H Minus First Strand cDNA Synthesis Kit from Fermentas (St Leon-Rot, Germany): 1 µg of total RNA and 1 µl of random hexamer primer (0.2 µg/µl) were diluted in diethylpyrocarbonate (DEPC)-treated water to a total volume of 12 µl, incubated at 70 °C for 5 min and chilled on ice prior to adding 4 µl of 5 × reaction buffer, 0.5 µl of RNAse inhibitor (20 U/µl) and 2 µl of 10 mM dNTP mix. After incubation at 25 °C for 5 min, 1 µl of reverse transcriptase (200 U/µl) was added, and the mixture was incubated for 10 min under the same conditions and for 60 min at 42 °C, prior to stopping the reaction by heating at 70 °C for 10 min and chilling on ice. Quantitative PCR was performed in a LightCycler from Roche (Penzberg, Germany) using the QuantiTect SYBR Green PCR kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions with 4.5 µl of cDNA (diluted 1 : 10), 1 µl of primers (5 µM) and 5 µl of SYBR Green master mix. A pre-incubation for 15 min at 95 °C was followed by 55 amplification cycles: 10 s at 95 °C, 10 s at 55 °C and 10 s at 72 °C. The melting curve for PCR product analysis was determined by rapid cooling down from 95 °C to 65 °C, and incubation at 65 °C for 15 s prior to heating to 95 °C. To normalize for equal mRNA/cDNA amounts, PCR reactions with VEGF-specific and with actin-specific primer sets were always run in parallel for each sample, and VEGF levels were determined according to the formula: 2^CP(VEGF)/2^CP(actin), where CP is the cycle number at the crossing point (0.3).

Nuclease protection assay and siRNA quantification

A synthetic DNA probe complementary to the VEGF siRNA anti-sense strand was 5′-end labeled with γ-³²[P] ATP (GE Healthcare, Milwaukee, WI, USA) using polynucleotide kinase (New England Biolabs, Beverly, MA, USA) according to the manufacturer's instructions. The probe sequence was: VEGF, 5′-GTGAATGCAGACCAAAGAAAGTCTTctt-3′. Seven micrograms of total tumor RNA (or GFP siRNA as a negative control) + 50 µg of Escherichia coli tRNA was hybridized in solution with 4 ng of radioactive probe. Following digestion with S1 nuclease at 42 °C for 20 min, samples were loaded on denaturing 12% acrylamide gels. Gels were exposed to a phosphoimager screen and analysed on a Typhoon 9200 instrument (GE Healthcare) with ³²[P]-labeled siRNA serving as standard for quantification.

VEGF ELISA

Human VEGF concentrations were determined from conditioned medium and from tumor lysates using the DuoSet ELISA Development System from R&D Systems (Abingdon, UK). For the preparation of tumor lysates, samples were homogenized in 500 µl of phosphate-buffered saline (PBS) using a Dounce homogenizer, or frozen in liquid nitrogen, ground in a mortar and suspended in 500 µl of PBS and, after centrifugation at 16 000 g supernatants were transferred to a fresh tube. Samples were diluted in reagent diluent [1% bovine serum albumin (BSA) in PBS], and the ELISA was performed according to the manufacturer's protocol (R&D Systems). Absorbance was measured in an ELISA reader at 450 nm with background subtraction at 595 nm. Recombinant human VEGF (R&D Systems) in appropriate buffers served as the standard.

Determination of serum levels of liver enzymes [γ-glutamyl transpeptidase (GGT), alkaline phosphatase (AP)] and of markers of immunostimulation [tumor necrosis factor (TNF)-α, interferon (IFN)-γ]

Serum levels of murine cytokines TNF-α and IFN-γ were determined using ELISA Development Kits from Preprotech (Rocky Hill, NJ, USA) in accordance with manufacturer's instructions. Athymic nude mice Hsd : Athymic Nude-Foxn1^nu or C57BL/6 wild-type mice were treated with PEI F25-LMW/siRNA complexes or 50 µg [intraperitoneal (i.p.)] of lipopolysaccharides (LPS from E. coli; Sigma, St Louis, MO) and compared with untreated mice.

Similarly, serum samples of the different mice were analysed for liver enzymes AP and GGT, using enzyme assays from DiaSys (Holzheim, Germany) according to the manufacturer's instructions. Briefly, enzyme activities were determined spectrophotometrically from the substrate turnover and enzyme units were determined according to the formulas provided.

In vivo siRNA tissue distribution and complex toxicity

For the radioactive determination of siRNA tissue distribution, 0.6 µg (0.05 nmol) of VEGF-specific siRNAs were [³²P] end-labeled at both strands using T 4 polynucleotide kinase and γ-[³²P] ATP prior to purification by microspin columns (Bio-Rad, München, Germany) to remove unbound radioactivity and complexation as described above. 3 × 10⁶ PANC-1 or PC-3 cells in 150 µl of PBS were injected s.c. into both flanks of athymic nude mice (nu/nu) and grown until they reached a size of approximately 8 mm in diameter. The complexes, or noncomplexed labeled siRNAs as a negative control, were dissolved in 200 µl of PBS for i.p. or s.c. injection, or in 100 µl of PBS for intravenous (i.v.) or intratumoral (i.t.) injection. At the time points indicated, mice were sacrificed, and samples, organs and tissues were removed. From different animals, equal amounts of any organ were taken and subjected to RNA preparation as described above. The total RNA was dissolved in 200 µl of DEPC-treated water, and 10 µl-samples were mixed with loading buffer, heat-denatured and subjected to agarose gel electrophoresis prior to blotting and autoradiography (Biomax, Eastman-Kodak, Rochester, NY, USA). Quantification was performed by phosphor imager analysis.

For toxicity studies, nontumor bearing mice were i.p., s.c., or i.v. injected with various amounts of PEI F25-LMW/siRNA complexes or PEI F25-LMW alone. Lethality, weight loss and local effects (formation of a scurf or wound, or changes in colour or macroscopic integrity at the site of injection), as well as any other signs of acute toxicity (sedation, visible impairment of the mice), were monitored hourly within the first 8 h and then daily for a period of at least 8 days. All animal studies have been approved by the Regierungspräsidium Giessen.

PEI F25-LMW/siRNA and Bevacizumab treatment in s.c. tumor models

4 × 10⁶ PANC-1 or 3 × 10⁶ PC-3 cells in 150 µl of PBS were injected s.c. into both flanks of athymic nude mice (Hsd : Athymic Nude-Foxn1^nu). When solid tumors were established, mice were randomized into treatment groups. Treatment was performed by i.p. or i.v. injection of 0.77 nmol (10 µg) of PEI F25-LMW-complexed specific or PEI-complexed nonspecific siRNA, and/or with 100 µg of i.p. injected Bevacizumab, at the time points indicated. Tumor volumes were monitored every 2–3 days, as indicated, and, upon termination of the experiment, mice were sacrificed and tumors removed. Pieces of each s.c. tumor were immediately fixed in 10% paraformaldehyde for paraffin embedding or snap-frozen for RNA preparation or ELISA (see above).

Immunohistochemistry

Immunostaining of paraffin sections of the tumors was performed essentially as described previously 42. Briefly, after deparaffinization with xylene and rehydration with graded alcohols, sections were incubated in 10 mM citrate buffer (pH 7.4) at 90 °C for 15 min and endogenous peroxidases were inactivated with 0.3% hydrogen peroxide at 4 °C for 30 min. After blocking with 10% normal horse serum in phosphate-buffered saline with Tween-20 (PBST)/2% BSA for 1 h at room temperature, the slides were incubated with mouse monoclonal anti-proliferating cell nuclear antigen (PCNA) antibodies (Dako, Hamburg, Germany) or mouse monoclonal anti-von Willebrand factor antibodies (Boehringer Mannheim, Germany), diluted 1 : 200 in PBST, at room temperature overnight in a wet chamber. For detection, biotinylated goat anti-(mouse-IgG) antibody (dilution 1 : 500; Vector Laboratories, Burlingame, CA, USA) was applied for 1 h, and immunoreactivity on the sections was visualized using a streptavidin-biotin-peroxidase complex (ABC kit, Vector Laboratories) according to the manufacturer's instructions and revealed with the peroxidase substrate diaminobenzidine (brown). When possible based on tumor size, each tumor was represented by at least two different sections originating from distant areas of the tumor mass. For the assessment of proliferation or blood vessel densities, blood vessels were counted or the percentage of proliferating cells was assessed in at least five fields per section.

Statistical analysis

Statistical analyses were performed by Student'st-test, one-way analysis of variance (ANOVA)/Tukey's multiple comparison test or two-way ANOVA using GraphPad Prism4, with significance levels as indicated. In therapeutic in vivo experiments, n represents the number of tumors.

Results

PEI/siRNA-mediated VEGF down-regulation in vitro

To assess the targeting efficacy of the PEI F25-LMW-complexed, chemically modified VEGF siRNAs, PANC-1 pancreas carcinoma cells were treated with PEI/siRNA4371 or PEI/siRNA4332 (both VEGF-specific) or with PEI/siRNA5246, containing an unrelated siRNA, as a negative control (for structures, see Figure 1). Complex sizes had been determined previously by atomic force microscopy as being in the range 60–130 nm, and laser doppler anemometry (LDA) had revealed zeta potentials in the range 35–40 mV (data not shown). The unrelated siRNA was chosen based on the absence of sequence homologies to the target gene or other human genes.

Figure 1. Design of the chemically modified siRNAs (a) and chemical structure of the modified nucleotides (b)

The analysis of cellular mRNA levels by quantitative real-time RT-PCR (qRT-PCR) revealed a profound approximately 65–85% down-regulation of VEGF mRNA levels, 72 h after treatment with 100 pmol/ml PEI/VEGF siRNA (Figure 2a). This effect was dependent on the dose of PEI/VEGF siRNA, was not associated with any visible signs of toxicity and was time-dependent, reaching maximum targeting efficacies after 3 days (data not shown). The mRNA down-regulation translated into reduced VEGF protein levels, as determined by ELISA from the tissue culture supernatant, because VEGF is a secreted protein. Because the medium was conditioned over a longer time period, it can be expected that time points with sub-optimal targeting efficacies (i.e. still too early and/or already too late for maximum knockdown) were also included and, therefore, that the maximum knockdown effect is somewhat under-estimated. Nevertheless, approximately 50% reduced VEGF concentrations were observed (Figure 2b). An assay of anchorage-dependent cell proliferation also revealed that this reduction in VEGF expression resulted in an approximately 50% decrease in proliferation (Figure 2c). It should be noted that VEGF medium levels in Figure 2b were normalized for cell numbers to avoid differences that are only based on different cell numbers at the time point of medium harvesting. Essentially, both VEGF-specific siRNAs displayed comparable efficacies on the level of mRNA, protein and cellular proliferation (note the absence of statistically significant differences in Figures 2a to 2c), and siRNA 4371 was chosen for subsequent experiments. Similalrly, in PC-3 prostate carcinoma cells, treatment with PEI F25-LMW-complexed VEGF siRNA 4371 led to the down-regulation of VEGF expression as determined by qRT-PCR (Figure 2d), ELISA (Figure 2e) and a proliferation assay (Figure 2f). The targeting efficacy, which was again compared with the nonspecific PEI/siRNA5246 as the negative control, was even more profound on protein levels and resulted in the near absence of cell proliferation under 2% serum conditions (Figure 2f). The comparison with a PEI F25-LMW-complexed, nonmodified, VEGF-specific siRNA revealed identical anti-proliferative effects (Figure 2g). A soft agar assay of anchorage-independent proliferation, which more closely resembles the in vivo situation, further confirmed the inhibition of proliferation with an approximately 50% reduction of colony formation upon treatment of cells with PEI F25-LMW-complexed VEGF siRNA 4371 (Figure 2h).

Figure 2. PEI F25-LMW/siRNA-mediated RNA interference in PANC-1 (a–c) and PC-3 cells (d–h) in vitro. Treatment of cells with complexes comprising of specific siRNAs 4371 or 4332 leads to a robust reduction of VEGF expression on mRNA (quantitative RT-PCR; a, d) and protein level (ELISA; b, e) compared to complexes with the nonspecific control siRNA 5246. This translates into anti-proliferative effects as determined in anchorage-dependent proliferation (d, f), which are identical when comparing PEI F25-LMW-complexed modified siRNA 4371 and a PEI F25-LMW-complexed nonmodified VEGF-specific siRNA (g; for additional clarity, a higher magnification is shown). Anti-proliferative effects are confirmed in anchorage-independent soft agar assays (h). Right: representative microscopic fields. A soft agar assay was not performed for PANC-1 cells because of poor colony formation of this cell line. *p < 0.05, **p < 0.01, ***p < 0.001, #, not significant

In vivo toxicity of PEI F25-LMW complexes and free PEI F25-LMW

As a pre-requisite for any treatment experiments, we assessed the in vivo toxicity of PEI F25-LMW/siRNA complexes and PEI F25-LMW alone. Intratumoral injection into s.c. PC-3 prostate carcinoma xenografts did not allow the administration of larger volumes as a result of the rather poor penetration into the tissue, caused pain to the animal, and represented a local rather than systemic application and, thus, was not pursued further. For systemic application, various amounts of complexes were administered by i.p., s.c. or i.v. injections into athymic nude mice at days 1, 3 and 5, and mice were analysed for 8 days as shown in Figure 3 (see also the Supporting information, Figure S1). For i.p. and i.v. injection, doses of 4 mg/kg PEI/0.4 mg/kg siRNA complexes did not reveal any signs of toxicity (lethality, weight loss, local effects such as formation of a scab or wound, or changes in colour or macroscopic integrity at the site of injection), sedation or other visible impairments of the mice). Subcutaneous injection of the same amounts, however, resulted in sedation/visible behavioural impairment (one of 10), weight loss (10 of 10) and scurf formation (eight of 10). All side-effects were reversible after a few hours (sedation/behavioural impairment) to a few days (weight loss; Figure 3b); however, s.c. injection did not show any obvious advantages, for example, regarding biodistribution (data not shown). On the basis of the absence of any signs of toxicity after i.p. and i.v. injections of 4 mg/kg PEI/0.4 mg/kg siRNA, amounts were increased three-fold. In the case of i.v. injection, this led to reversible weight loss (three of three) and sedation (one of three). By contrast, no adverse effects were observed after i.p. injection, and only amounts ten-fold above the intended treatment doses resulted in toxicity.

Figure 3. Toxicity profile of PEI F25-LMW/siRNA complexes in vivo upon various modes of administration. (a) Various amounts of complexes were administered i.p., s.c., i.v. or i.t. into s.c. PC-3 prostate carcinoma xenografts) injections into athymic nude mice at days 1, 3 and 5 (see upper panel for details), and mice were analysed over 8 days for lethality, acute side-effects (sedation or other visible impairments of the mice), weight loss and local effects (formation of a scurf or wound (b, right), or changes in colour or macroscopic integrity) at the site of injection visible from the outside. Weight loss upon s.c. injection of a complex comprising of 4 mg/kg PEI F25-LMW/0.4 mg/kg siRNA was completely reversible within approximately 4 days after treatment (b, left)

The possibility of the complete dissociation of PEI F25-LMW/siRNA complexes upon administration in vivo also requires the separate evaluation of the toxicities of the individual components. While the modified siRNAs have been shown previously not to cause any toxic effects at the concentrations used in the present study 43, free PEI F25-LMW as a charged polymer may well exert toxicity. Consequently, for the two relevant modes of administration (i.e. i.p. and i.v. injection), PEI F25-LMW was tested up to the amounts to be applied in the PEI F25-LMW/siRNA complexes, thus monitoring toxicity based on the assumption of a 100% complex disruption. No signs of toxicity were observed at either of these amounts (12 mg/kg for i.p. and 4 mg/kg for i.v. injection) or with lower amounts (see Supporting information, Figure S1).

In vivo tissue distribution experiments

For successful gene targeting, the efficient delivery of intact siRNA molecules into the target tissue/target organ is critical. Therefore, we determined the tissue distribution of naked or PEI F25-LMW-complexed siRNA molecules. For detection, siRNAs were [³²P]-end labeled and analysed by tissue RNA preparation, gel electrophoresis, blotting and autoradiography. Thus, bands only represent fully intact, labeled siRNA molecules, whereas free nucleotides or free label, arising from nucleolytic degradation or phosphatase-mediated de-labeling, are purified off and siRNAs are not visible without label. Thirty minutes after i.p. injection of PEI F25-LMW-complexed, chemically modified siRNA, strong signals were observed in tumor and muscle, somewhat weaker bands in samples from kidney and lung, and weak signals in the liver and brain. Levels of intact siRNA in the blood were barely detectable, indicating that the signals in any organ truly represent siRNA taken up into the tissue without interference of siRNA in the residual blood in the nonperfused organ (Figure 4a, upper left). By contrast, upon i.p. injection of the same amount of naked, chemically modified siRNA, no signals were detected, except for a very weak band in muscle (Figure 4a, upper right). Thus, despite the chemical modifications of siRNA molecules introduced to increase their halflives in the blood, tissue delivery of intact siRNAs was completely dependent on the PEI F25-LMW complexation. Comparably high siRNA levels in the liver were also observed after i.v. injection of PEI F25-LMW-complexed siRNA, with this mode of administration resulting in a totally different siRNA tissue distribution pattern (Figure 4a, lower left). More specifically, a strong signal was also observed in the lung, whereas band intensities in muscle and kidney were low, and no siRNA was detected in the brain and in the tumors. Notably, the siRNA band in the blood indicated the presence of intact siRNA in the circulation for at least 30 min after i.v. injection. Again, injection of naked siRNA did not result in siRNA delivery, with the exception of a weak band in the lung (Figure 4a, lower right).

Figure 4. (a) Biodistribution of intact full-length siRNA upon various modes of administration. SiRNA molecules were [³²P]-end labeled and detected in tissue lysates by gel electrophoresis, blotting and autoradiography after i.p. or i.v. injection of naked or PEI F25-LMW-complexed siRNA at 30 min as indicated. (b) Quantification of the biodistribution of [³²P]-end labeled siRNA upon a single i.p. (left) or i.v. (right) injection. Values are given in percentage of the total intact siRNA/mg tissue at the time point of harvesting the mice (60 min). (c) Biodistribution is comparable for a PEI F25-LMW-complexed, unmodified siRNA versus PEI F25-LMW-complexed, modified siRNA 4371

These data on PEI-mediated siRNA delivery were extended by the direct and quantitative comparison of the biodistribution profile between i.p. and i.v. injection after 60 min (Figure 4b). For better comparison, the percentages of residual siRNA per mg tissue were determined. In the case of i.p. injection, strong signals were observed for kidney, liver and tumor, and weaker signals were observed for lung, muscle and brain. By far the highest uptake, however, was found in the spleen. The very efficient siRNA delivery into this organ was also observed upon i.v. injection of PEI F25-LMW/siRNA complexes. Comparable levels were observed in the liver, whereas signals were much weaker in lung and kidney and very low in muscle, brain and, notably, in tumor. Similarly, strong signals were detected 4 h after i.p. injection (not shown) with a tissue distribution pattern different from the 30-min time point. More specifically, apart from tumor and muscle, very strong signals were now also detected in kidney and in liver, with the latter showing the most profound difference between 30 min and 4 h. Furthermore, signal intensities were slightly less in lung and brain, and no siRNA band was detected in the blood (Figure 4b). Finally, a PEI-complexed, unmodified siRNA and the PEI-complexed, chemically modified siRNA 4371 were compared. Quantification of radioactive siRNA as above revealed identical results with regard to tissue siRNA levels in various organs (Figure 4c).

Taken together, these experiments show that the mode of administration of PEI F25-LMW/siRNA complexes, but not the chemical composition of the siRNAs, profoundly determines the tissue distribution of siRNAs, and suggests that i.p. injection allows the efficient delivery of siRNAs into s.c. tumor xenografts.

Anti-tumor effects of PEI F25-LMW/VEGF siRNA complexes upon systemic treatment of tumor-bearing mice are synergistic with Bevacizumab treatment in s.c. prostate carcinoma xenografts

In the first set of experiments, PC-3 cells were injected s.c. and upon formation of small s.c. tumors as clearly visible nodules, mice were treated with i.p. injections of PEI F25-LMW-complexed specific (4371) or unrelated (5246) siRNA (10 µg) at the time points detailed in Figure 5. Tumors in the untreated and in the control treated group grew well and at identical rates. By contrast, approximately 10 days after the start of the treatment, reduced tumor sizes were observed in the PEI F25-LMW/siRNA 4371 group, indicating an anti-tumorigenic effect. Here as well as in subsequent experiments, no animals died during the treatment, and no other adverse effects were observed. For ethical reasons, mice in the control groups had to be sacrificed at day 35 as a result of tumor-induced weight loss and obvious suffering from the tumor burden. To immunohistochemically analyse and compare the tumors between the two groups, mice in the F25-LMW/siRNA 4371 treatment group were sacrificed at the same time. At this time point, tumor masses in the treatment group were approximately 65% of the control group (p < 0.001). The treatment with Bevacizumab was largely comparable to the PEI F25-LMW/VEGF siRNA group, with only that at day 35 appearing to be somewhat more efficient.

Figure 5. Synergistic anti-tumorigenic effects of VEGF inhibition in s.c. PC-3 prostate carcinoma xenografts upon PEI F25-LMW/siRNA-mediated RNA interference and Bevacizumab treatment. Upon formation of visible s.c. tumors, mice were randomly assigned into five groups with six mice each and i.p. injected with PEI F25-LMW-complexed VEGF-specific (4371, red) or nonspecific (5246, black) siRNA (10 µg), or with Bevacizumab (green) at the time points indicated in the treatment schedule (upper panel; black boxes refer to all treatment groups, white boxes indicate the treatment being performed only in the ‘Bevacizumab’ and ‘Bevacizumab + 4371’ groups), or left untreated (blue). The strongest anti-tumor effects were observed upon double treatment with PEI F25-LMW/VEGF siRNA and Bevacizumab (yellow). Right: representative examples of mice treated with PEI F25-LMW-complexed specific or nonspecific siRNA at the time point of termination of the experiment. p-values indicate statistically significant differences of the data points compared to control groups: **p < 0.01, ***p < 0.001

Most interesting, however, was the combination of both treatments which led to synergistic effects. The onset of anti-tumorigenic effects was earlier (i.e. already 7 days after start of the treatment) and tumor growth inhibition was by far more profound than the individual effects of Bevacizumab or PEI F25-LMW/VEGF siRNA. This was true for the time point of termination of the experiment in the control group (day 35) with an 80% tumor growth reduction in the double treatment group compared to control, as well as for a prolonged period of time afterwards. In addition, although tumors in the Bevacizumab group almost reached the size of the control tumors at the time point of termination, tumors in the double-treatment group were still less than half the size and showed profoundly inhibited growth (Figure 5). Thus, we conclude that PC-3 tumor xenograft treatment with PEI F25-LMW/VEGF siRNA and with Bevacizumab exerts additive or synergistic, strongly anti-tumorigenic effects.

Quantification of PEI F25-LMW-mediated siRNA delivery into s.c. prostate carcinoma xenografts and analysis of tumors

To precisely determine the amounts of intact siRNA molecules in the s.c. tumors after the whole treatment period, we performed an RNAse protection assay-based method of siRNA detection 44, 45, which did not require the administration of radioactively labeled siRNAs. Total RNA from tumors was isolated and hybridized with radioactive probes specific for the VEGF siRNA or the control siRNA, respectively. For the determination of absolute amounts, siRNA standards with known concentrations were quantified in parallel. The comparison between the F25-LMW/siRNA 4371 and the F25-LMW/siRNA 5246 groups revealed no cross-reactivity of the probes, thus indicating their high specificity. From the siRNA standards, the limit of detection was determined to be below 2 fmol (Figure 6a). Total siRNA amounts varied between the tumors and were determined to be in the range 2900–17 000 pg/tumor (Figure 6b). The average value of 10 000 pg/tumor indicated that approximately 0.1% of the siRNA from a single injection could still be found in each tumor tissue at the time point of harvesting. This number, however, does not include the siRNA molecules already excreted or still remaining in the peritoneum, and so may rather under-estimate the total siRNA amount reaching the tumor tissue.

Figure 6. Quantification of intact siRNA molecules by RNase protection assay in tumor samples of mice treated with PEI F25-LMW/siRNA complexes. (a) Total RNA from tumors was isolated and hybridized in solution with a probe specific for the VEGF siRNA or the control siRNA, respectively. No cross-reactivity of the probes is observed, thus indicating their high specificity (right: different tumor samples are presented) and based on siRNA standards (left panel) the limit of detection is below 2 fmol. (b) Total siRNA amounts vary between the tumors (quantification of the different tumor samples shown above) and were determined to be in the range 2900–17 000 pg/tumor

Analysis of the tumors at the time points of termination of the experiment revealed a marked 70–80% down-regulation of VEGF mRNA levels upon i.p. treatment with PEI F25-LMW/VEGF siRNA (Figure 7a). This effect was observed in the single treatment group, as well as upon combination treatment with Bevacizumab, whereas the antibody alone did not exert any effects on VEGF mRNA levels. The PEI F25-LMW/VEGF siRNA-mediated down-regulation of VEGF mRNA translated into a reduction of VEGF protein levels in the tumors as determined by ELISA from tumor lysates (Figure 7b). This reduction, however, was less pronounced (approximately 35%) than mRNA knockdown and was stronger after Bevacizumab treatment, which resulted in approximately 20% residual VEGF protein levels. Serum VEGF levels were below the limit of detection in all experiments and thus did not allow the determination of targeting efficacies (data not shown). The monitoring of the mouse body weight revealed a slight decline over the whole period as a result of the tumor burden in all groups, but no differences between the groups (Figure 7c), thus confirming the earlier results regarding the absence of toxicity (see above). Additionally, the immunohistochemical analysis of the tumors also demonstrated an anti-proliferative effect of the i.p. F25-LMW/siRNA 4371 treatment, as determined by a more than 50% reduction in the number of proliferating tumor cells compared to the control treatment group (Figure 7d). Similarly, the staining for blood vessels showed an approximately 50% reduction of tumor blood vessel density in the i.p. F25-LMW/siRNA 4371 treatment group. Although the staining was generally weak and blood vessel density was rather heterogeneous between different areas within a section, the quantification was reliable enough to indicate an anti-angiogenic effect (Figure 7d).

Figure 7. Analysis of prostate carcinoma xenografts in mice treated with i.p. injection of PEI F25-LMW/siRNA and Bevacizumab for (a) VEGF mRNA and (b) VEGF protein levels. (c) Mouse body weights are unaffected by the treatments and only decline as a result of the increased tumor burden. (d) PEI F25-LMW/VEGF siRNA treatment of the mice inhibits in vivo tumor cell proliferation (left) and blood vessel density (centre), as determined by immunohistochemistry for PCNA and von Willebrand factor (right: representative examples of brown staining). Scale bar = 20 µm. **p < 0.01, ***p < 0.001, #, not significant

Further analysis of the efficacy of siRNA delivery by a very sensitive, modified RNAse protection assay revealed, in accordance with the results obtained from the radioactive tissue distribution assay (Figure 4b), no detectable siRNA amounts upon i.v. injection of PEI F25-LMW/siRNA complexes (see Supporting information, Figure S2a). Concomitantly, the treatment of tumor-bearing mice via i.v. injection of PEI F25-LMW-complexed, VEGF-specific siRNA did not result in tumor growth reduction and, upon termination of the experiment, no VEGF down-regulation on the mRNA or protein level was detected in the tumors (data not shown). By contrast, a single i.p. injection already led to measurable siRNA levels (see Supporting information, Figure S2b, left), which were further enhanced upon multiple PEI F25-LMW/siRNA injections (see Supporting information, Figure S2b, right).

Absence of nonspecific hepatoxotic or immunostimulatory effects

To assess in more detail the safety of our PEI/siRNA particles with regard to unwanted other effects such as immunogenic reactions or impaired organ function, we analysed serum levels of two established markers of liver function/liver damage, GGT and AP, and of two established markers of immunostimulation, the pro-inflammatory cytokines TNF-α and IFN-γ. To this end, we also included immunocompetent mice. Levels of liver enzymes were comparable between immunocompetent and immunocompromised mice. More importantly, no substantial increase in GGT or AP levels was observed upon repeated treatment with PEI/siRNA complexes (see Supporting information, Figure S3a).

Although the siRNAs had been tested for the absence of immunostimulatory activity (data not shown), we analysed the absence of immunostimulation upon PEI formulation and in vivo application. As shown in the Supporting information (Figure S3b, left), no increase in TNF-α levels was observed. This was true for athymic nude mice as well as for immunocompetent mice. Furthermore, we determined IFN-γ levels. Because IFN-γ is produced in T cells that are absent in athymic nude mice, we focused on immunocompetent mice to exclude false-negative results. As shown in the Supporting information (Figure S3b, right), no increase in IFN-γ levels was observed upon PEI/siRNA treatment. We conclude from these experiments that, beyond the toxicity data shown in Figure 3, PEI F25-LMW/siRNA complexes are safe with regard to (the absence of) immunostimulation and liver damage.

Anti-tumor effects of PEI F25-LMW/VEGF siRNA complexes are comparable to Bevacizumab treatment in s.c. pancreatic carcinoma xenografts

As a second tumor model, the anti-tumorigenic effects of PEI F25-LMW/siRNA-mediated VEGF targeting and Bevacizumab treatment were tested in s.c. PANC-1 pancreas carcinoma xenografts. Intraperitoneal treatment with PEI F25-LMW/VEGF siRNA alone was already sufficient to produce a profound (approximately 65%) reduction in tumor growth compared to the control group (see Supporting information, Figure S4a). Similarly, Bevacizumab resulted in a very similar growth inhibition, indicating that both treatments were again comparable regarding their efficacy, which was markedly higher than that in the PC-3 prostate carcinoma xenografts. Concomitantly, no additional anti-tumorigenic effects were observed upon combination of both treatment strategies (see Supporting information, Figure S4a). At the time point of termination of the experiment, the analysis of the tumors by the RNAse protection assay again demonstrated the presence of intact siRNA in most tumors at approximately 60–380 pg/tumor, with some samples being below the limit of detection (data not shown). Furthermore, quantitative RT-PCR and ELISA revealed a PEI F25-LMW/siRNA-mediated reduction of VEGF on mRNA as well as protein levels (see Supporting information, Figure S4b, upper panel). Again, no effects of PEI/VEGF siRNA and/or Bevacizumab treatment on mouse body weight (see Supporting information, Figure S4b, lower panel) or other unwanted effects were observed.

Discussion

The poor prognosis of advanced pancreatic and prostate carcinoma, as well as the limited treatment options, emphasize the need for novel therapeutic strategies. In addition to cytostatic drugs, inhibitory antibodies or small molecule inhibitors, the discovery of RNAi has extended the portfolio towards highly specific gene-targeting approaches. However, this requires the efficient systemic delivery of RNAi-inducing agents such as siRNAs, based on the generation and in vivo testing of appropriate delivery vehicles and their optimal modes of administration.

PEIs were introduced previously as a delivery platform for siRNAs in vitro and in vivo19. Although several, if not all, PEIs are able to complex and protect siRNAs, efficient PEI/siRNA-mediated gene targeting in vivo relies on the stability of the complexes including that in the presence of serum (i.e. little aggregation or extracellular complex disruption), their penetration into tissues, the efficient cellular uptake of the complexes, as well as the subsequent intracellular release of the siRNA. Additional issues are the biocompatibility of the complexes, as well as of the single components, also upon repeated treatments. In the present study, we demonstrate the in vivo efficacy and safety of branched 4–10 kDa polyethylenimine PEI F25-LMW 37 and, for PEI-mediated delivery, we employed for the first time chemically modified siRNA. The analysis of in vitro targeting efficacies and the in vivo delivery (as reported in the present study as well as unpublished data) reveals that the chemical modification neither impairs nor enhances PEI-mediated siRNA delivery and targeting efficacy, which rather relies on the PEI as a delivery vehicle than the introduction of chemical siRNA modifications.

In both tumor models, the comparison between PEI F25-LMW/siRNA-mediated VEGF targeting and the treatment with Bevacizumab shows similar anti-tumorigenic efficacies. This is despite their completely different mechanisms of action at transcriptional and post-translational levels, respectively. Thus, it can be assumed that a combination may enhance the anti-tumorigenic efficacies exerted through VEGF blockade based on different mechanisms of action, as observed in the present study with regard to the synergistic effects in prostate carcinoma xenografts. This may be relevant with regard to avoiding excess siRNAs, which can compete for the intracellular RNAi machinery as suggested previously 46, and may allow the enhancement of individual drug effects and/or the reduction of therapeutically relevant doses. The latter aspect is particularly important when aiming to avoid nonspecific siRNA side-effects exerted, for example, through the activation of toll-like receptors of the innate immune system, which have been shown to be dependent on siRNA concentrations and to increase upon siRNA delivery in liposomal formulations 47, 48. In this context, it should also be noted that we always employed relatively low siRNA amounts compared to other studies that relied on large siRNA quantities.

Combinations may also include PEI/siRNA-mediated VEGF targeting and established cytostatics, as described previously for Bevacizumab 14, 15. With regard to optimal therapeutic regimens, however, it should also be taken into consideration that anti-angiogenic effects may impair the delivery of other anti-cancer drugs as a result of the reduction of tumor vasculature, or may, in contrast, transiently increase drug delivery as a result of the ‘normalization’ of the abnormal structure and function of tumor vasculature 49. This emphasizes the need for optimized treatment protocols.

The RNase protection assay proved to be a very sensitive and accurate method for quantifying full-length siRNAs, and is therefore superior to other protocols relying on labeled siRNA and the detection of either the label alone (without assessing siRNA integrity or the presence of free label) or the use of radioactively labeled siRNAs. Stronger anti-tumorigenic effects are observed in pancreatic tumor compared to the prostate carcinoma xenografts, despite less siRNA delivery and less VEGF reduction, and rather low initial VEGF expression levels in PANC-1 cells 6. This also indicates that profound biological effects of RNAi may not necessarily rely on particularly high levels of the target gene overexpression or on maximum targeting efficacies. It should also be noted that the VEGF-specific siRNAs used in the present study are also able to target mouse VEGF, which parallels a therapeutic setting in humans where a distinction between tumor and stroma VEGF would neither be made nor wanted. Consequently, the determination of tumor VEGF levels by the ELISA specific for human VEGF may underestimate the total VEGF knockdown. Finally, the differences in VEGF protein levels between the PEI/siRNA and PEI/siRNA + Avastin treatment groups were less prominent than the effects on prostate carcinoma xenograft growth. Although this finding could reflect the already very low VEGF levels in the Avastin treatment group, making it difficult to detect further reduced levels with sufficient accuracy, it should also be noted that RNAi-mediated knockdown of VEGF expression leads to reduced VEGF levels already in the target cells. Because VEGF is able to exert a ‘direct’ proliferative effect on tumor cells, as shown in the present study as well as previously 10–13, the reduced VEGF expression close to its site of action may exert (anti-tumoral) effects that are stronger than expected from the overall decrease of VEGF in the whole tumor mass.

Our qualitative and quantitative assessment of the biodistribution profiles in various organs and the siRNA uptake into the tumors further reveals that the mode of administration is of critical importance. Indeed, i.v. injection leads to poor siRNA levels in the tumors and, concomitantly, to the absence of anti-tumorigenic effects. By contrast, i.p. injection, which is already a relevant administration route in tumor therapy, results in both siRNA uptake and a profound inhibition of tumor growth. This also demonstrates the direct correlation between siRNA levels and VEGF down-regulation, and further excludes nonspecific effects of the PEI or the siRNA with regard to the reduction of tumor growth. A ‘depot effect’ may be beneficial for i.p. administration leading to prolonged release of partially aggregated complexes from the site of injection, thus increasing the time period of siRNA delivery into the tumor. For i.v. application, PEI with modifications to increase the complex circulation half-lives and/or to increase tumor-specific uptake may be required. Nevertheless, the nonmodified PEI F25-LMW introduced in the present study for in vivo use, already demonstrates good results with regard to efficacy and biocompatibility. The toxicity data obtained reveal no side-effects of the complexes or the free carrier at the concentrations used. This is also true for the absence of hepatotoxic or immunostimulatory effects of the PEI/siRNA complexes. In proof-of-principle studies with regard to the relevance of a given gene product, direct injection into the tumor may represent an alternative, reasonable mode of administration. However, because of other therapeutic options, including surgery, this approach would be of little therapeutic value and would be limited to a few tumors not accessible for surgery. Therefore, we did not pursue this further. It should also be noted that PEI F25-LMW/siRNA complexes can be stored in the freezer for several months with full retention of their bioactivity 38. Thus, in contrast to many other delivery platforms, this provides a standardized, ready-to-use formulation that will be particularly important for therapeutic applications. Taken together, PEI F25-LMW/siRNA complexes represent efficient and safe tools for VEGF depletion in vivo and show great promise in comparison to, or in addition to, conventional treatment with Bevacizumab.

Acknowledgements

This work was supported by grants from the Deutsche Forschergruppe Forschungsgemeinschaft/German Research Foundation (Forschergruppe/Research Group 627 ‘Nanohale’, AI 24/6-1 to S.H., F.C. and A.A.) and the Deutsche Krebshilfe/German Cancer Aid to A.A. We are grateful to Helga Radler and Andrea Wüstenhagen for their expert technical assistance.