Adeno-Associated Virus Vector-Mediated Systemic Delivery of IFN-β Combined with Low-Dose Cyclophosphamide Affects Tumor Regression in Murine Neuroblastoma Models

Purpose: Type I IFNs (IFN-α/β) have shown significant antitumor activity in preclinical models but limited efficacy and significant toxicity in clinical trials. We hypothesized that the antitumor activity of type I IFNs could be enhanced by chronic, low-dose systemic delivery and sought to test this in murine neuroblastoma models.

Experimental Design: Continuous liver-generated expression of human IFN-β (hINF-β) was achieved through a gene therapy–mediated approach using adeno-associated virus vectors encoding hIFN-β (AAV hINF-β). Orthotopic localized retroperitoneal and disseminated models of neuroblastoma were established using three different xenografts. Immunohistochemical analysis and ELISA were used to evaluate the antiangiogenic effect of therapy.

Results: The development of both localized orthotopic (retroperitoneal) and disseminated neuroblastoma was prevented in all mice expressing hINF-β. Continued growth of established retroperitoneal tumors, treated with AAV hINF-β as monotherapy, was significantly restricted, and survival for mice with established, disseminated disease was significantly prolonged following administration of AAV hINF-β. Analysis of treated tumors revealed a significant antiangiogenic effect. Mean intratumoral vessel density was diminished and expression of the angiogenic factors vascular endothelial growth factor and basic fibroblast growth factor were both decreased. Finally, combination therapy in which AAV hIFN-β was used together with low-dose cyclophosphamide resulted in regression of both established retroperitoneal and disseminated disease.

Conclusions: AAV-mediated delivery of hIFN-β when used as monotherapy was able to restrict neuroblastoma growth due in part to inhibition of angiogenesis. When used in combination with conventional chemotherapy, AAV hIFN-β was able to effect complete tumor regression.

Neuroblastoma is the most common solid extracranial malignancy of childhood and is responsible for ∼15% of all pediatric cancer deaths (1). “High-risk” disease is rarely curable, even with current multimodal treatment approaches that include increasingly aggressive myeloablative chemoradiotherapy. Clearly, new treatment strategies are needed for children with this disease. One promising agent is IFN. Since their identification in the late 1950s as potent antiviral agents (2), type I IFNs (α/β) have become increasingly recognized for their antitumor activity (3–5). The pleiotropic antitumor mechanisms of action of type I IFNs include direct tumor cell cytotoxicity (6) and indirect activity through immunomodulation (7) and inhibition of angiogenesis (8–12). There is evidence to suggest that neuroblastoma is likely susceptible to each of these antitumor mechanisms. We have shown that recombinant IFN can have a direct antitumor effect on neuroblastoma xenografts (13). Other preclinical studies have suggested that neuroblasts are susceptible to cytotoxic immune effector mechanisms both in vitro and in vivo (14). Finally, additional preclinical studies have shown that neuroblastoma is also susceptible to angiogenesis inhibition through a variety of agents, including TNP-470 (15–17), vascular endothelial growth factor (VEGF)-Trap (18), a truncated soluble form of the vascular endothelial growth factor receptor-2 (19, 20), and pigment epithelium-derived factor (21, 22). In addition, standard chemotherapeutic agents, when given using a frequent, low-dose schedule, seem capable of treating tumors that had previously been resistant to those agents by destroying the neovascularity required by a progressing tumor (23, 24).

Despite exciting preclinical results, however, the antitumor efficacy of type I IFNs in clinical trials has been limited (25). Significant contributing factors include the very short half-life of type I IFNs, making effective dosing problematic, and their significant toxicity at high doses (26). An alternative dosing schedule in which there is continuous, low-dose delivery might avoid the toxicity of IFN while maintaining its antitumor efficacy. In addition, two general concepts have emerged regarding angiogenesis inhibition as an anticancer strategy. First, the traditional scheduling of conventional cytotoxic therapy, administration of a maximum tolerated dose followed by a recovery period, may permit the microvascular cells in a tumor bed, with their slower rate of cell division, to recover and proliferate, thereby continuing to provide neovasculature to support tumor regrowth (23). A metronomic schedule in which cytotoxic drugs are given more frequently, or even continuously, without a long treatment-free interval is an attractive alternative. This strategy may be more effective at controlling tumor progression, at a lower total dose, even if the tumor cells are resistant to the drug, by targeting the endothelial cells (23, 27). In addition, toxicity with this schedule may be decreased as lower peak circulating levels of protein are generated. Second, long-term delivery of angiogenesis inhibitors is likely to be important because treated tumor sites, although small and dormant, may maintain their capacity for growth, invasion, and metastasis if released from the restrictions of angiogenesis inhibition (28).

An alternative to chronic administration of recombinant proteins is the use of a gene therapy approach in which genetic material encoding the therapeutic protein is transferred to a host target cell where it mediates long-term protein expression. Gene therapy strategies for inhibition of tumor-associated angiogenesis, delivering a variety of different inhibitors, including type I IFNs, have already been tested in a number of different murine tumor models, with some success (29–32). However, most of these studies have used either naked DNA, or retroviral or adenoviral vectors. Unfortunately, retroviral vectors may not be suitable for human use and the transfer of naked DNA is typically an inefficient, transient process. Adenoviral-mediated gene transfer is also complicated by transient transgene expression as well as a host immune response to transduced target cells.

Adeno-associated virus (AAV) is a nonpathogenic, helper-dependent member of the parvovirus family. A number of properties make AAV-based vectors promising for antiangiogenic anticancer gene therapy (33). Most importantly, unlike the other gene delivery systems just mentioned, recombinant AAV (rAAV) vectors have been shown to direct long-term transgene expression from nondividing cells (34, 35). In addition, these vectors have an excellent safety profile. In addition, unlike other vectors of viral origin, AAV has never been associated with any human disease and is naturally replication deficient, thereby providing an added measure of safety. In addition, rAAV is nonimmunogenic; the genes encoding wild-type viral proteins have been removed thus reducing the potential for invoking a cell-mediated immune response due to the expression of foreign viral proteins. We hypothesized that the antiangiogenic and antitumor activity of IFN-α/β could be enhanced by chronic, low-dose delivery mediated by AAV and sought to test this in murine neuroblastoma models.

Cell lines. The human neuroblastoma cell lines IMR-32 and SK-N-AS, purchased from American Type Culture Collection (Manassas, VA), and NB-1691, provided by Dr. P. Houghton (Memphis, TN) were maintained in RPMI 1640 (Bio-Whittaker, Walkersville, MD) supplemented with 10% heat-inactivated fetal bovine serum (Mediatech Cellgro, Herndon, VA), 100 units/mL penicillin and 100 μg/mL streptomycin (Life Technologies, Grand Island, NY), and 2 mmol/L l-glutamine (Life Technologies). 293T cells (human embryonic kidney cells expressing SV40 large T antigen, American Type Culture Collection) were maintained in DMEM (Mediatech Cellgro) supplemented with penicillin-streptomycin and l-glutamine as above.

Effects of IFN-β in vitro. Tumor cells were plated at a density of ∼3 × 10⁵ cells per well. The following day, different doses of recombinant human IFN-β (hIFN-β, Avonex, Biogen, Inc., Cambridge, MA) were added to the culture medium. The culture medium was aspirated and the cells were refed with hIFN-β containing medium each day. Cell counts, Annexin V fluorescence-activated cell sorting, and cell cycle analysis were done on the tumor cells after 96 hours of culture with recombinant hIFN-β. All experiments were done in triplicate.

Recombinant AAV vector construction, production, and purification. Construction of the pAV5 and pAV2 CAG hIFN-β and CAG hFIX vector plasmids have been described previously (13). These vector plasmids each include the CMV-IE enhancer, β-actin promoter, a chicken β-actin/rabbit β globin composite intron, and a rabbit β globin polyadenylation signal (CAG) mediating the expression of the cDNA for hIFN-β or human clotting factor IX. The expression cassettes are flanked either by the AAV-5 or AAV-2 inverted terminal repeats. The hIFN-β cDNA was purchased from InvivoGen (San Diego, CA). The hFIX cDNA was a gift from Prof. G. Brownlee (Oxford). Construction of the pAV2 CMV luciferase vector plasmid has also been described previously (36). rAAV vectors were made by the transient transfection method described previously (37). The necessary AAV5 vector plasmid containing the type 5 inverted terminal repeats (pAAV5-7D05) and packaging plasmid (pAAV5-2) were provided by Dr. R. Kotin (NIH; ref. 38). rAAV5 virions were purified using mucin affinity column chromatography (39). rAAV vectors pseudotyped with serotype 8 capsid were generated by transient transfection of 293T cells using one of the vector plasmids, a second plasmid, pAAV8-2 (40) from Dr. J. Wilson (Philadelphia, PA), for packaging and a third plasmid, pHGTI-adeno1, to provide adenoviral helper function. The plasmid HGTI-adeno1 contains adenovirus serotype 5 genomic sequences (Genbank accession no. M73260, coordinates 9840-13258, 21442-28139, and 30819-35810) cloned into LITMUS39. This plasmid recapitulates the structure of the helper plasmid XX6 (41), without the adenoviral inverted terminal repeats and with better plasmid yields in Escherichia coli.⁵

These rAAV-2/8 vectors were purified using our recently described ion exchange chromatography method (42). Standard slot-blot analysis was used to determine the vector particle titer. The vector stocks were consistently free of contamination with wild-type AAV and cellular and adenoviral proteins as judged by our previous described methods (43). rAAV vectors pseudotyped with capsid 8 were used exclusively once they became available because of the significantly greater transduction efficiency of the liver with vectors of this pseudotype than with rAAV-5 (44).

Murine tumor models. Retroperitoneal (orthotopic) xenografts were established in C.B-17 severe combined immunodeficient mice (The Jackson Laboratory, Bar Harbor, ME) by injection of 1.5 × 10⁶ neuroblastoma tumor cells in 150 μL PBS behind the left adrenal gland via a left subcostal incision during administration of 2% isofluorane. Measurements of the retroperitoneal tumors were done by ultrasonography. At the conclusion of each experiment, mice were sacrificed and portions of their tumors were snap-frozen in liquid nitrogen or fixed in formalin. Disseminated neuroblastoma was established by injecting 1.5 × 10⁶ NB-1691 tumor cells via tail vein. This reliably creates disseminated disease, primarily in the liver, spleen, adrenal glands, kidneys, retroperitoneal lymph nodes, and bone marrow (45). AAV vector particles were also given via tail vein, in a volume of 250 μL PBS. Cyclophosphamide (Bristol-Myers Squibb Co., Princeton, NJ) was given at a dose of 160 mg/kg s.c. every 6 days for two to four doses. All murine experiments were done in accordance with a protocol approved by the Institutional Animal Care and Use Committee of St. Jude Children's Research Hospital.

Bone marrow analysis. Bone marrow was harvested from the femurs and tibias of mice following sacrifice. Fluorescence-activated cell sorting analysis for CD9 (DAKO, Carpinteria, CA) and CD56 (BD Biosciences, San Jose, CA) double-positivity was used to detect the presence of neuroblastoma cells in the bone marrow. This analysis has a sensitivity of 1:10⁴ cells.

Human IFN-β immunoassay. Quantitation of AAV-mediated hIFN-β expression in mouse plasma was done using a commercially available sandwich immunoassay (ELISA, PBL Biomedical, Brunswick, NJ).

Protein extraction and measurement of angiogenic factors. Protein lysates were made by homogenizing snap-frozen tumor specimens with a Dounce (Kontes, Vineland, NJ) homogenizer in 1 mL of lysis buffer. The homogenates were incubated on ice for 30 minutes and centrifuged at 10,000 × g for 10 minutes at 4°C. The supernatants were then recentrifuged, collected, and frozen at −80°C for later use. Total protein was quantified for each sample using the Bradford assay. The levels of intratumoral basic fibroblast growth factor (bFGF) and VEGF protein were analyzed by Quantikine ELISA kits for the respective proangiogenic factor (R&D Systems, Minneapolis, MN). The minimal detectable levels of protein with these assays were bFGF (3 pg/mL) and VEGF (5 pg/mL).

Natural killer cell depletion. To deplete mice of natural killer (NK) cells, 100 μg of rat anti-mouse CD122 (BD Biosciences) were given by i.p. injection on the day of AAV hIFN-β administration. This dose was repeated 5 days later.

Tumor immunohistochemistry and terminal deoxynucleotidyl transferase–mediated nick-end labeling assay. Formalin-fixed, paraffin-embedded 5-μm tumor specimens were analyzed by immunohistochemistry with an anti-CD34 antibody to determine endothelial cell density as previously described (46). Endothelial cell density in hotspots was determined with light microscopy at 400× by the method described previously by Weidner et al. (47). Apoptosis in s.c. and retroperitoneal tumors was determined by the terminal deoxynucleotidyl transferase–mediated nick-end labeling method using a commercially available in situ apoptosis detection kit (Serologicals, Norcross, GA). Densities of apoptotic cells were determined by 400× light microscopy in the field with the highest density of apoptotic cells in a region that had no evidence of necrosis. The detection of microscopic disease was facilitated by staining sections with an antibody for human chromogranin (Chemicon, Temecula, CA).

Statistical analyses. Results are reported as means ± SE. The Sigmaplot program (SPSS, Inc., Chicago, IL) was used to analyze and graphically present the data. An unpaired Student's t test was used to analyze statistical differences among in vitro cell densities, final xenograft weights and volumes, tumor angiogenic protein expression, intratumoral microvessel density, and apoptosis. Differences in survival were compared by ANOVA. P < 0.05 was considered statistically significant.

Sensitivity of neuroblastoma cell lines to recombinant IFN-β in vitro. To determine the direct effects of hIFN-β on the human neuroblastoma cell line NB-1691, these cells were cultured for 96 hours in medium containing varying doses (100-10,000 IU/mL) of recombinant hIFN-β or vehicle control. The effect of IFN-β was evaluated by measuring changes in cell proliferation, degree of apoptosis, and cell cycle profile. NB-1691 cells had only a modest degree of sensitivity to the direct effects of recombinant hIFN-β in vitro. Cell counts after 96 hours of culture with 10,000 IU/mL of recombinant hIFN-β were 63% (P < 0.005) of control-treated cells. The decreased proliferation of these cells did not seem mediated through cell cycle arrest but a 30% increase in apoptosis as assessed by Annexin V. In comparison, >99% of the highly sensitive leukemia cell line SEM were apoptotic following similar in vitro treatment with hIFN-β.⁶

When IMR-32 and SK-N-AS cells were treated in vitro with hIFN-β, no significant antiproliferative or direct cytotoxic effect was seen at any dose.

Prevention of tumor engraftment by adeno-associated virus–mediated IFN-β delivery. The first set of in vivo studies evaluated the ability of AAV-mediated, continuously delivered hIFN-β to prevent human neuroblastoma tumor cell engraftment in immunodeficient mice (see Table 1). Four cohorts of mice were used in the retroperitoneal tumor model. Each received either a different dose of AAV5 hIFN-β (5 × 10¹¹, 2.5 × 10¹¹, or 1 × 10¹¹ vector particles per mouse) or 5 × 10¹¹ vector particles of AAV hFIX (control vector). Three weeks following tail vein administration of the rAAV vectors, systemic levels of hIFN-β were 341 ± 104 pg/mL (n = 4), 75 ± 28 pg/mL (n = 7), and 20 ± 20 pg/mL (n = 4), respectively. Mice that received control vector had hIFN-β levels of <10 pg/mL (n = 6). NB-1691 cells were then injected into the retroperitoneal space. No gross or microscopic evidence of tumor engraftment could be detected in the retroperitoneum of any of the mice treated with AAV hIFN-β at any of the three doses at necropsy 5 weeks after tumor cell inoculation. Mice treated with AAV hFIX all had tumors that weighed an average of 2.92 ± 0.35 g at that time.

The ability of hIFN-β to prevent disseminated disease was also evaluated (see Table 1). Two different doses of AAV5 hIFN-β were used (9 × 10¹¹ or 4.5 × 10¹¹ vector particles per mouse; n = 5 per group). Control mice were given 9 × 10¹¹ vector particles of AAV hFIX. No overt, vector-related toxicity (death and weight loss) from this dose of vector was observed in any of the mice that received treatment or control vector. Three weeks following tail vein administration of the rAAV vectors, systemic hIFN-β levels of 1,448 ± 306 and 534 ± 211 pg/mL, respectively, were confirmed in the first two cohorts of mice. NB-1691 cells were then injected into each of the mice via tail vein. No gross or microscopic evidence of tumor engraftment could be detected in the liver or retroperitoneum of any of the mice treated with AAV hIFN-β at necropsy 7 weeks after tumor cell inoculation. In addition, none of the 10 mice had disease detected in the bone marrow by fluorescence-activated cell sorting for CD9 and CD56. Control mice treated with AAV hFIX each had miliary disease in the liver and extensive retroperitoneal lymphadenopathy. In addition, two of four control vector–treated mice evaluated had evidence of neuroblastoma in the bone marrow.

Stabilization of established tumors with adeno-associated virus vector–mediated IFN-β delivery. We next sought to test the antitumor efficacy of AAV hIFN-β in a more rigorous and relevant model, the treatment of established disease. Retroperitoneal NB-1691 tumors were established in male C.B-17 severe combined immunodeficient mice. When the tumors were an average of 0.4 ± 0.1 cm³, 3 weeks after tumor cell injection, cohorts containing five mice each were given either 1 × 10¹⁰ or 6 × 10¹⁰ AAV2/8 hIFN-β vector particles or 6 × 10¹⁰ control vector particles via tail vein. The vector dose for these mice was lower than had been used previously in the prevention model because of the switch to AAV pseudotyped with the more efficient serotype 8 capsid. Eleven days later, all mice were sacrificed and their tumor size evaluated (see Fig. 1A). Mice that had received the control vector, AAV2/8 hFIX, had an average tumor volume of 9.36 ± 0.6 cm³. Mice that had received the higher dose of AAV hIFN-β (mean expression at sacrifice = 810 ± 63 pg/mL) had an average tumor volume of 2.53 ± 0.6 cm³ (P < 0.003, compared with control), whereas mice that had received the lower dose of AAV hIFN-β (mean expression at sacrifice = 353 ± 63 pg/mL) had an average tumor volume of 4.70 ± 0.7 cm³ (P < 0.03, compared with control). An additional cohort of five mice had significantly larger tumors (average volume = 1.06 ± 0.37 cm³) at the time of AAV hIFN-β administration (6 × 10¹⁰ vector particles per mouse). Mean hIFN-β levels in these mice at sacrifice 11 days later was 720 ± 292 pg/mL and the mean tumor volume was 4.76 ± 0.7 cm³. Thus, growth of both relatively small and large retroperitoneal tumors was significantly restricted by AAV hIFN-β monotherapy. To confirm that the primary tissue transduced with systemically given AAV vector was the liver and not the tumor itself, additional tumor-bearing mice were given 6 × 10¹⁰ AAV2/8 cytomegalovirus luciferase vector particles per mouse via tail vein. Fourteen days later, photon emission from successfully transduced cells was detected using the Xenogen Bioluminescence imaging system (Xenogen Corp., Stanford, CA; ref. 48) only from the liver and not the retroperitoneal tumor (see Fig. 1B).

Although these experiments were done in severe combined immunodeficient mice, these mice still retain some degree of NK cell activity. To exclude any component of immunomodulation by IFN-β, a subset of experiments were repeated in which severe combined immunodeficient mice were depleted of NK cells by treating them with CD-122-depleting antibody during the course of treatment of established retroperitoneal disease. No significant difference in tumor volume was observed between tumors in mice treated with AAV hIFN-β with or without NK cell depletion (data not shown).

The effect of continuous delivery of hIFN-β on mice with disseminated neuroblastoma was then evaluated. Twenty mice were given NB-1691 cells via tail vein. Three weeks later, four of these mice were sacrificed to confirm the presence of established disease. Each of these four mice had gross evidence of disease in the liver and retroperitoneum and two had evidence of bone marrow metastases by fluorescence-activated cell sorting. Eight of the remaining 16 mice were then given 6 × 10¹⁰ AAV2/8 hIFN-β vector particles via tail vein; the other eight received a similar dose of control vector. Expression of hIFN-β in the mice treated with AAV hIFN-β 4 weeks after vector administration was an average of 420 ± 115 pg/mL. Survival following vector administration for the mice treated with AAV hIFN-β was significantly longer than for those treated with control vector (57.0 ± 12.1 versus 10.3 ± 0.3 days, P < 0.012; Fig. 1C). However, despite persistent expression of hIFN-β, these mice did eventually die of disease.

Regression of established retroperitoneal tumors with combination therapy. A conventional cytotoxic chemotherapeutic agent, cyclophosphamide, was used in combination with AAV hIFN-β in an effort to effect regression of established neuroblastoma rather than to simply restrict further growth. Cyclophosphamide was given in an every sixth day dosing schedule that has been shown to be antiangiogenic (23). Four weeks after 20 mice had been given retroperitoneal NB-1691 tumor cells, ultrasonography was done to confirm the presence of established tumors. The mean tumor volume at that time was 0.7 ± 0.2 cm³. The mice were then divided into four treatment groups such that there was no significant difference in tumor volume at the time of therapy initiation. Cohorts containing five mice then each received either AAV2/8 hIFN-β (6 × 10¹⁰ vector particles) alone, cyclophosphamide alone, control vector alone, or the combination of AAV2/8 hIFN-β and cyclophosphamide. Eight days after vector injection, average systemic levels of hIFN-β of 1,031 ± 213 and 794 ± 126 pg/mL were detected in combination therapy– and AAV hIFN-β alone–treated mice, respectively. Mice were sacrificed 2 weeks after the initiation of therapy. Their tumors were measured and harvested for histology and protein analysis. Combination therapy–treated tumors were significantly smaller than control tumors (0.07 ± 0.04 versus 5.50 ± 0.9 cm³, P < 0.0001) or those treated with either AAV hIFN-β alone (1.19 ± 0.05 cm³, P < 0.004) or cyclophosphamide alone (1.14 ± 0.5 cm³, P < 0.006; see Fig. 2A). In fact, in three of five mice treated with combination, there was no gross evidence of residual tumor.

The differences in tumor size were reflected in increasing degrees of apoptosis of the tumor cells (see Fig. 2B). Control vector–treated tumors had significantly fewer terminal deoxynucleotidyl transferase–mediated nick-end labeling–positive apoptotic tumor cells (counts per high power field) than tumors in any of the treatment groups (37 ± 2, P < 0.001, combination therapy; 30 ± 3, P < 0.007, cyclophosphamide alone; 31 ± 3, P < 0.009, AAV hIFN-β; 13 ± 4, control vector alone). The mean intratumoral vessel density of these treated tumors was also evaluated in an effort to assess the contribution of the antiangiogenic activity of hIFN-β on the antitumor effectiveness of this treatment approach. Immunohistochemical analysis of sections from tumors harvested from treated mice revealed a significantly lower mean intratumoral vessel density (counts per high power field) in the combination treatment group compared with that of the control vector–treated group (26 ± 7, P < 0.004, combination therapy; 36 ± 2, cyclophosphamide alone; 34 ± 3, AAV hIFN-β; 45 ± 3, control vector alone; see Fig. 2C). The cause for this decrease in mean vessel density was further investigated by evaluating the level of expression of two important angiogenic factors, VEGF and bFGF, in tumors of each of the treatment groups. This was done by ELISA evaluation of protein extracted from harvested tumors. Levels of expression of both VEGF and bFGF were significantly decreased in tumors that received combination therapy compared with control vector–treated tumors (see Fig. 2D). VEGF expression in combination therapy–treated tumors was 30 ± 16 pg/100 μg total protein (P < 0.03, compared with control), 75 ± 7 pg/100 μg total protein in AAV hIFN-β alone–treated tumors, and 47 ± 5 pg/100 μg total protein in cyclophosphamide alone–treated tumors compared with 182 ± 67 pg/100 μg total protein in control vector–treated tumors. This represented declines of VEGF expression of 84%, 59%, and 74%, respectively. Similar results were found for the levels of bFGF expression. bFGF expression was 18 ± 10 pg/100 μg total protein in combination therapy–treated tumors (P < 0.001, compared with control), 259 ± 60 pg/100 μg in AAV hIFN-β alone–treated tumors, and 311 ± 70 pg/100 μg in cyclophosphamide alone–treated tumors compared with 433 ± 57 pg/100 μg in control vector–treated tumors. This represented declines of bFGF expression of 96%, 40%, and 28%, respectively.

Five additional mice were treated with combination therapy that was initiated when the mean tumor volume was again ∼0.70 cm³ by ultrasound. These tumors treated with combination therapy showed regression in all mice, as determined by ultrasound, and by day 32 had regressed completely. These mice, that were treated with only four doses of cyclophosphamide but that had stable levels of hIFN-β expression, were sacrificed 5 months following tumor cell challenge (4 months following AAV hIFN-β administration and 3 months following completion of cyclophosphamide). None of these five mice had gross or microscopic evidence of persistent retroperitoneal disease.

Finally, combination therapy was used to treat established retroperitoneal tumors generated from other human neuroblastoma cell lines to show that this treatment strategy might be broadly applicable in the treatment of neuroblastoma. Combination therapy was initiated for established retroperitoneal IMR-32 tumors and SK-N-AS tumors when they were average sizes of 0.09 ± 0.01 cm³ (day 35) and 0.13 ± 0.03 cm³ (day 14), respectively, as determined by ultrasonography. After baseline volume measurements were obtained, half of the mice (n = 5) with each tumor were given 6 × 10¹⁰ vector particles AAV2/8 hIFN-β and the other half were given a similar dose of AAV2/8 hFIX. Two weeks after AAV injection and after having received only two doses of cyclophosphamide, all mice were sacrificed. Mice that had received the control vector all had very large tumors (IMR-32 average volume = 2.33 ± 0.14 cm³, SK-N-AS = 7.34 ± 1.55 cm³). None of the mice that had received IMR-32 cells and AAV2/8 hIFN-β had gross or microscopic evidence of tumor. Two of the mice with established SK-N-AS treated with AAV2/8 hIFN-β also had no evidence of residual disease. Interestingly, in each of the other three mice, a small necrotic, cystic lesion was found in the retroperitoneum, which on histologic examination, was found to have foci of microscopic, residual disease (Fig. 3).

Improved survival from disseminated neuroblastoma treated with combination therapy. In the final set of experiments, the ability of combination therapy to treat established disseminated disease was evaluated. Treatment of mice was initiated 5 weeks following tail vein inoculation of NB-1691 cells. In addition to having groups treated with control vector and cyclophosphamide and AAV hIFN-β alone (4 × 10¹⁰ vector particles per mouse), four cohorts of mice were treated with cyclophosphamide together with decreasing doses of AAV hIFN-β (4 × 10¹⁰, 2 × 10¹⁰, 4 × 10⁹, and 2 × 10⁹ vector particles per mouse). Five weeks later, hIFN-β expression levels in these mice were 1,443 ± 93, 515 ± 88, 125 ± 65, and 45 ± 16 pg/mL, respectively. Control mice survived an average of an additional 14.0 ± 2.1 days and those treated with cyclophosphamide alone survived an average of 59.4 ± 0.9 days. The mice treated with AAV hIFN-β alone (mean hIFN-β expression at 5 weeks = 1,645 ± 166 pg/mL) survived an average of 57.0 ± 12.1 days (P < 0.0003, compared with control vector–treated mice). Each group treated with combination therapy survived significantly longer than mice that had received control vector (53.0 ± 18.5, 84.4 ± 6.4, 106.0 ± 12.3, and 113.8 ± 17.9 days, respectively), with mice that received the three highest doses of AAV hIFN-β, together with cyclophosphamide, having a significantly longer survival than those treated with either cyclophosphamide or AAV hIFN-β alone (P < 0.05; Fig. 4). Dose-dependent differences in survival were seen in mice treated with AAV hIFN-β, when used together with a standard dose of cyclophosphamide. Despite maintaining stable levels of hIFN-β 3 months after vector administration, all mice eventually died of disease relapse.

Because disseminated, high-risk neuroblastoma is a very difficult therapeutic challenge, clinical trials are ongoing in which novel treatment approaches are being evaluated, including immunotherapy, radionuclide therapy, the use of agents that induce tumor apoptosis or differentiation, and angiogenesis inhibition. In this study, we have shown that chronic AAV-mediated delivery of hIFN-β can successfully prevent localized and disseminated neuroblastoma engraftment and significantly retard the growth of established retroperitoneal and disseminated disease when used as monotherapy. When used in combination with a standard cytotoxic chemotherapeutic agent, cyclophosphamide, that was delivered infrequently and in low doses, hIFN-β was able to effect regression of established localized and disseminated disease.

IFNs, multifunctional regulatory cytokines that control cell function and regulation, are among the most commonly studied agents for the biological therapy of cancer. The antitumor efficacy of IFN is based on its pleiotropic effects. The direct antiproliferative effects (49) are often assumed to be the primary mechanism involved in their efficacy as anticancer agents. However, IFNs have other antineoplastic effects, including immunomodulation, through T and NK cell stimulation (50) regulation of cellular responses to inducers of apoptosis including bcl and TRAIL (51) and inhibition of angiogenesis (8). Their antiangiogenic properties seem mediated in part by their ability to down-regulate expression of tumor-elaborated proangiogenic factors including bFGF (9), VEGF (12), and matrix metalloproteinase-9 (9, 52). Clinical experience has proven the antiangiogenic potential of type I IFNs in the treatment of a variety of pediatric vascular neoplasms including hemangiomas, malignant hemangiopericytoma, and pulmonary hemangiomatosis (53). This antiangiogenic capacity of IFN seems to have contributed significantly in the murine neuroblastoma models used in this study. Down-regulation of two of the principal proangiogenic factors, VEGF and bFGF, in response to hIFN-β therapy, likely resulted in the decreased mean vessel density of treated tumors, thus contributing to the increased tumor cell apoptosis and restriction in tumor growth. It is unlikely that the immunomodulatory activity of IFN-β contributed significantly to the anticancer effectiveness of our treatment approach as the experiments were all done in immunodeficient mice with any residual NK activity being blocked by a depleting antibody. The relative contribution of the direct, tumoricidal effect of hIFN-β is somewhat less certain. The NB-1691 neuroblastoma cells showed a modest degree of sensitivity to hIFN-β in vitro; thus, it is likely that a direct cytotoxic effect on the tumor cells also contributed to the antitumor effect in vivo. This is probably particularly true in the experimental setting of preventing tumor cell engraftment, because small (<2 mm³) tumors may not even require angiogenesis to survive; yet, no evidence of even microscopic disease was detected in this setting. However, neither the SK-N-AS nor the IMR-32 cells showed susceptibility to the direct cytotoxic effects of hIFN-β in vitro, suggesting an even more prominent role for the antiangiogenic activity of hIFN-β in effecting regression of these tumors. This would further suggest that this combination therapy, in addition to being effective against neuroblastoma, might be broadly applicable to a variety of solid tumors, which are not necessarily sensitive to the direct effects of IFN-β but which require new blood vessel formation to grow. The antiangiogenic effect seems to have been quite rapid as antitumor efficacy in treating established tumors was often seen within 2 weeks of AAV hIFN-β administration. This is particularly noteworthy when recognizing that even with the use of serotype 8 vectors that mediate transgene expression fairly rapidly from transduced target cells, 3 to 5 days are required following vector administration to achieve appreciable levels of hIFN-β expression (44).

A gene therapy–mediated approach to the delivery of hIFN-β, as well as other angiogenesis inhibitors, is attractive because continuous, low-level expression of these proteins, as would be generated from gene-modified cells, may be the optimal delivery schedule, achieving therapeutic efficacy while minimizing toxicity. Type I IFNs are currently the most commonly used cytokine in human patients (7). Although these cytokines have great promise for the treatment of other solid neoplasms, clinical trials have often failed to show efficacy or have been limited by systemic side effects from the high-dose administration which was thought to be required for effective therapy (25). Pharmacokinetic studies have shown that daily i.v. and s.c. dosing are associated with a systemic half-life of <5 hours (26), perhaps limiting efficacy by decreasing exposure to sites of primary tumor and metastatic implants. Furthermore, other studies have shown that optimization of the dose and frequency of IFN delivery is necessary for maximal antitumor efficacy (9). It may be that continuous, low-dose gene therapy–mediated delivery may address the limitations of current IFN clinical protocols, while avoiding some of the systemic side effects seen when administering IFN on a high dose, bolus schedule. Others have successfully used adenoviral vectors for effecting delivery of hIFN-β as therapy in preclinical tumor models (29–31). However, adenovirus typically generates only transient levels of transgene expression (2-3 weeks) and may elicit a host immune response that may have significant untoward consequences in immunocompetent hosts. rAAV seems an ideal gene delivery system as it can mediate long-term transgene expression without generating a significant host immune response because no viral proteins are expressed.

Prior studies have shown that synergistic antitumor efficacy between angiogenesis inhibitors and conventional chemotherapeutic agents can be achieved, especially when the cytotoxic drugs are delivered with an antiangiogenic dosing schedule (27, 54–56). We have found this to be true in our murine neuroblastoma model in which AAV hIFN-β monotherapy restricted the growth of established disease and when used in combination with conventional cytotoxic chemotherapy, was able to effect regression of established disease. However, following cessation of cyclophosphamide, but with continued delivery of hIFN-β, these mice eventually succumbed to relapsed neuroblastoma. The immunohistochemical analysis shown in Fig. 3 may help to explain this finding as it showed that despite the significant antitumor efficacy of AAV hIFN-β, microscopic disease may still remain. We are currently evaluating the mechanism for this apparent resistance of the tumor cells and determining whether resuming administration of cyclophosphamide would be able to establish a second remission of disease. In addition, we are evaluating the use of regulatable expression cassettes to determine whether intermittent expression of hIFN-β would prevent the development of apparent resistance.

Dose-dependent differences were seen when treating established local and disseminated disease. However, the determination of an “optimal” dose could not be established. The kinetics of AAV-mediated transgene expression are sigmoidal rather than linear, in relating the dose of vector given and level of transgene expression established. Thus, determining the lowest dose at which neuroblastoma engraftment could be prevented was difficult because vector doses below 2 × 10⁹ vector particles per mouse did not consistently give detectable systemic levels of IFN-β expression. Conversely, higher vector doses, as might be used to improve the results for survival studies, did not consistently give higher systemic levels of IFN-β expression.

Neuroblastomas grow quickly, are highly vascularized, and metastasize early; hence, inhibition of angiogenesis may be effective for its treatment. An attractive new approach for the treatment of neuroblastoma, with its angiogenic phenotype, particularly in high-risk disease, is long-term gene therapy–mediated delivery of angiogenesis inhibitors. rAAV, because of its ability to safely mediate long-term transgene expression, and IFN, because of its ability to inhibit endothelial activation as well as to cause direct tumor cell cytotoxicity and stimulate the immune system, are a logical gene delivery system and angiogenesis inhibitor to use. Consideration of this approach should be given for the treatment of patients with neuroblastoma, perhaps in combination with cytotoxic therapy, and in the setting of minimal residual disease.

We thank Dorothy Bush for her assistance with immunohistochemistry, Stacey Glass for her assistance with ultrasonography, and the staff of the Vector Core Facility at St. Jude Children's Research Hospital for their assistance in generating pseudotyped rAAV vectors particles required for this study.