Continuous Delivery of IFN-β Promotes Sustained Maturation of Intratumoral Vasculature

Type I IFNs (IFN-α/β) have shown significant antitumor activity in animal models (1-3) through a variety of mechanisms, including direct tumor cell cytotoxicity (4), immunomodulation (5), and inhibition of angiogenesis (6-9). However, despite promising preclinical results, the antitumor efficacy of type I IFNs in treating human cancer has been disappointing (10). The relative inactivity of IFN in clinical trials might be explained, at least in part, by its rapid clearance, with the distribution half life in human subjects being ∼5 min (11). Attempts at delivering larger doses of IFN to maintain plasma levels have been limited by significant systemic toxicity (10).

Continuous delivery of type I IFNs might be a more efficacious dosing regimen while also potentially avoiding some of the systemic side effects associated with high-dose bolus administration. This type of pharmacokinetic profile can be achieved most readily using a gene therapy approach in which gene-modified host cells mediate continuous protein synthesis for systemic delivery. In addition, there are other, more general, potential benefits to gene therapy–mediated delivery of anticancer agents. (a) Chronic drug delivery might be achieved following a single intervention or administration. (b) Difficulties with protein production and maintenance of function, especially when “scaling up” for clinical trials, may be avoided by in situ expression in host tissues. (c) Continuous, low-level expression of these agents, as would be generated from gene-modified cells, may be the optimal delivery schedule. (d) Potential side effects of therapy might be avoided by either limited local expression or regulatable, intermittent, systemic expression, both of which can be achieved through gene therapy–mediated approaches.

Cancer gene therapy strategies are being tested in a number of different murine tumor models, with some success. Of the various vector delivery systems, a number of properties make adeno-associated virus (AAV) vectors among the most promising for cancer gene therapy (12). AAV is a nonpathogenic, helper-dependent member of the parvovirus family. Most importantly, unlike most other gene delivery systems, AAV vectors have been shown to direct long-term transgene expression from nondividing cells. In addition, these vectors have an excellent safety profile. 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. Recombinant AAV is nonimmunogenic; the wild-type viral genes have been removed, thus reducing the potential for evoking a cell-mediated immune response due to the expression of foreign viral proteins.

We hypothesized that the antitumor activity of IFN-β could be enhanced by its continuous delivery, mediated by AAV, and tested this strategy in different murine tumor models. Using this approach, we have identified an additional activity not previously ascribed to IFN-β, that being arterialization of the intratumoral vessels, a process characterized by remodeling and maturation of the vasculature of established tumors (13). We have further shown that this pharmacologic remodeling of the typically inefficient tumor vasculature improved intratumoral delivery of systemic chemotherapy.

Tumor cells were injected in either a heterotopic location, the subcutaneous space (U87 human glioma), or an orthotopic location, the retroperitoneal space (NB-1691 human neuroblastoma) of male C.B-17 severe combined immunodeficient mice. These cell lines were chosen because they are only moderately sensitive to the direct cytotoxic effects of IFN-β (14, 15). Treatment was initiated with administration of either AAV IFN-β or a control AAV encoding human clotting factor IX (FIX) via the tail vein, after established disease had been documented either by palpation (subcutaneous tumors) or by ultrasonography (retroperitoneal tumors). These AAV vectors were pseudotyped with serotype 8 capsid to efficiently target the liver and achieve rapid-onset, high-level transgene expression (16). Subcutaneous U87 gliomas were treated with 5 × 10¹⁰ AAV IFN-β vector particles administered via tail vein 18 days after tumor cell inoculation when tumors were of an average size of 0.25 ± 0.05 cm³. This vector dose led to systemic levels of human IFN-β (hIFN-β; 22.3 ± 3.4 ng/mL) by 5 days after vector administration. These levels remained stable over the remaining course of the experiment, being 24.5 ± 5.6 ng/mL at the time of mouse sacrifice 5 days later. This resulted in significant growth restriction of established subcutaneous U87 tumors as compared with those grown in mice that received the same dose of control vector (Fig. 1A).

Established retroperitoneal NB-1691 tumors were treated in a similar manner. An orthotopic tumor model was included because previous studies have suggested that tumor location can have a significant effect on the angiogenic profile of a growing tumor (17). To study the effect of treatment noninvasively and in real time, these tumors were followed by both serial ultrasonography and bioluminescence imaging, the NB-1691 cells having been modified to stably express luciferase (18). Treatment of established retroperitoneal NB-1691 tumors was initiated 17 days following tumor cell inoculation, when tumors were a mean size of 0.13 ± 0.03 cm³. Restriction in further growth of these tumors, as observed with U87 tumors, was again achieved with AAV IFN-β treatment (5 × 10¹⁰ vector particles per mouse; average plasma levels of hIFN-β: 19.5 ± 4.7 ng/mL at day 22, 27.3 ± 6.4 ng/mL at day 38; Fig. 1B). Hepatotoxicity with this approach was not observed. Serum levels of alanine aminotransferase measured 1, 3, and 5 weeks after administration of 5 × 10¹⁰ AAV8 vector particles encoding either hIFN-β or human FIX (hFIX) were always within normal limits (mean, 53.8 ± 17.8 units/L; reference range, 26-120 units/L).

We have previously shown that the bioluminescent signal (photons per second) in untreated retroperitoneal neuroblastoma xenografts is directly proportional to the tumor burden (NB-1691 cell number) in vitro and in vivo (18). It was therefore surprising to observe that the effect of AAV IFN-β monotherapy on established retroperitoneal neuroblastomas, as assessed by bioluminescent imaging, seemed to be counter to the assessment by ultrasonography. The bioluminescent signal of tumors treated with control vector declined somewhat with increasing tumor size after reaching a peak level (Fig. 1C, left), an observation previously made by others and felt to be due to hypoxia and necrosis within an expanding tumor mass (19). Interestingly, however, the bioluminescent signal of tumors treated with AAV IFN-β actually increased during the course of IFN therapy, although the average tumor size, as judged by ultrasonography, had stabilized (Fig. 1C, right). Because the tumor burden seemed to be constant and because IFN-β had no direct effect on luciferase expression by the tumor cells in vitro (data not shown), we hypothesized that the luminescent signal increased due to more efficient perfusion of the IFN-β–treated tumors, resulting in increased delivery of the luciferin substrate to the tumor cells and/or better oxygenation of these cells.

Improved tumor perfusion was confirmed using microbubble contrast-enhanced ultrasonography. We previously found that two variables were particularly useful for evaluating intratumoral perfusion using this technique: the change in contrast enhancement (in decibels) and the rate of this change (in decibels per second; ref. 20). Using this imaging modality, we determined that perfusion of retroperitoneal NB-1691 and subcutaneous U87 tumors was significantly greater and more homogeneous in AAV IFN-β–treated mice, as compared with mice given control vector (Fig. 2A and B). The functional consequence of the improved perfusion of these tumors was confirmed by showing a decrease in the degree of intratumoral hypoxia in size-matched tumors, as determined by staining for pimonidazole-protein adducts within tumor tissue sections (Fig. 2C and D).

Morphologic changes of the vasculature within IFN-β–treated tumors were assessed by intravital microscopy. This analysis revealed that tumors in mice treated with control vector were supplied by highly tortuous, disorganized vessels (Fig. 3A, left). A dramatic difference in vascular morphology was observed in IFN-β–treated mice, however. Vessels in these tumors displayed improved structural organization, wherein the vessels exhibited a more hierarchical organization, as seen in the microvascular networks of healthy tissues (Fig. 3A, right). Immunohistochemical analysis was done to evaluate the effect of continuous IFN-β therapy on the tumor vasculature at the cellular level. The most striking effect was a significant increase in the number of intratumoral perivascular cells, as identified by their positive α-smooth muscle actin (αSMA) staining. Both U87 and NB-1691 tumors in control vector–treated mice contained a paucity of these cells whereas tumors of both histologies in IFN-β–treated mice had an abundance of perivascular cells investing the small-diameter arterioles (Fig. 3B and C). The mean endothelial cell density, as assessed by CD34 staining, was unchanged in treated U87 tumors (14,816 ± 1,201 versus 15,005 ± 1,325 CD34⁺ pixels/tumor section, control versus IFN-β treated; P = 0.885) but actually increased in treated NB-1691 tumors (7,541 ± 892 versus 22,036 ± 1,547 CD34⁺ pixels/tumor section, control versus IFN-β treated; P = 0.001). Despite this, both U87 and NB-1691 tumors in IFN-β–treated mice exhibited a greater vessel maturity index, as defined by the ratio of perivascular cells (αSMA positivity) to endothelial cells [CD34 positivity; U87: 0.07 ± 0.02 versus 0.68 ± 0.09 (P < 0.02); NB-1691: 0.04 ± 0.01 versus 0.61 ± 0.07 (P < 0.001), control versus IFN-β treated]. To help distinguish perivascular cells from activated fibroblasts, both of which can be detected by αSMA staining, additional sections from control and IFN-β–treated tumors (four each) were stained with an anti–platelet-derived growth factor receptor β (PDGFRβ) antibody because PDGFRβ⁺ cells within tumors have previously been shown to be perivascular cells (21). PDGFRβ⁺ cells were nearly completely absent in untreated control tumors but were abundant and found to be lining tumor vessels in IFN-β–treated tumors (Fig. 3D)

In additional experiments, a dose-response relationship was shown between the systemic level of IFN-β expression and both αSMA staining (Fig. 3E, left) and tumor size (Fig. 3E, right). Cohorts of mice (n = 5 per group) bearing established retroperitoneal neuroblastomas were treated with either control vector (5 × 10¹⁰ vector particles per mouse) or decreasing doses of AAV IFN-β (5 × 10¹⁰, 5 × 10⁹, or 5 × 10⁸ vector particles per mouse). These doses of vector resulted in mean plasma levels of expression of 55.9 ± 3.6, 1.9 ± 0.4, and 0.05 ± 0.004 ng/mL, respectively. When measured by an antiviral assay, plasma from mice that received the high and intermediate doses of vector had activities of 1,500 ± 316.23 and 70 ± 19.74 IU/mL, respectively. No antiviral activity was detectable in plasma from mice that had received the lowest dose of AAV IFN-β vector.

The functional consequences of the increased coverage of tumor vessels with vascular smooth muscle cells were shown in additional experiments in which diminished permeability of the vessels (Fig. 4A) and consequent decreased tumor interstitial pressure (Fig. 4B) were shown in tumors grown in AAV IFN-β–treated mice. Finally, confocal images further revealed the pronounced difference in the vasculature of IFN-β–treated and control tumors, highlighting the physical relationship between endothelial cells and perivascular cells, with nearly complete lining of the IFN-β–treated tumor vessels with perivascular cells. The vessels in control tumors had almost no investment with perivascular cells (Fig. 4C). Thus, the morphologic, histologic, and functional consequences of continuously delivered IFN-β seemed to be a remodeling and maturation of the tumor vasculature with decreased vessel permeability, increased structural support provided by the increase in perivascular cells, and an overall increase in the efficiency of tumor perfusion.

One of the important theoretical consequences of the process of vascular remodeling and maturation is improved delivery of systemic chemotherapy to a tumor that is more efficiently perfused and oxygenated, with resulting improvement in the antitumor efficacy of these agents. We have previously shown antitumor synergy between AAV-mediated continuous systemic delivery of IFN-β and conventional cytotoxic chemotherapy (15, 22). We now sought to determine whether the in vivo effects correlated with improved drug delivery. To do this, mice bearing established retroperitoneal NB-1691 tumors (average size of ∼0.2 cm³) were first given AAV IFN-β vector (n = 8; 5 × 10¹⁰ vector particles per mouse via tail vein). Two weeks later, four of the eight mice given AAV IFN-β and four untreated control mice bearing size-matched tumors received a 2 g/kg bolus infusion of topotecan via tail vein. One hour later, blood was collected and tumors were harvested. Subsequent measurement of topotecan tumor penetration (ratio of tumor to plasma topotecan concentration) confirmed improved intratumoral delivery of systemically administered drug. Topotecan penetration in tumors grown in mice pretreated with AAV IFN-β was more than double that in tumors in mice that received control vector (54% versus 25%; tumor topotecan concentration: control, 11.1 ± 0.4 ng/mL; IFN-β, 15.8 ± 0.6 ng/mL; Fig. 5). To determine whether the effect of continuous IFN-β on the tumor vasculature was sustained, the other four mice that had been given AAV IFN-β did not receive the topotecan dose until 4 weeks after vector administration, 2 weeks later than in the first cohort. Sustained efficiency of delivery was confirmed, as topotecan penetration in tumors in these mice was not significantly different from the tumors in the mice that had received the topotecan dose sooner after the initiation of IFN-β therapy (48% versus 54%, respectively; tumor topotecan concentration: IFN-β, 14.5 ± 0.7 ng/mL; Fig. 5). Because we harvested the tumors 1 h after giving only a single dose of topotecan, not enough time had elapsed to enable us to determine whether the improved delivery affected tumor growth.

Because there is little cross-reactivity or signaling of the human cognate of IFN-β with the murine IFN receptor (23), changes in the intratumoral vasculature were not likely to have been mediated by a direct effect of systemic hIFN-β on the mouse vasculature. Therefore, in seeking to identify the mediator of hIFN-β–induced vascular maturation, we focused on identifying tumor-elaborated factors responsible for this effect. Angiopoietin-1 was selected as a candidate effector because the histologic and morphologic phenotype of the vasculature within the tumors treated with continuous hIFN-β seemed to be consistent with the role attributed to angiopoietin-1 in angiogenesis (24, 25) and was similar to that seen in other studies in which there was enforced overexpression of angiopoietin-1 by tumor xenografts (26, 27). Real-time PCR analysis of RNA extracted from U87 and NB-1691 tumor cells treated in vitro with recombinant hIFN-β revealed that, in fact, there was an associated increased expression of angiopoietin-1. The levels of angiopoietin-1 mRNA increased 2- to 4-fold when U87 and NB-1691 cells were treated with 10,000 IU/mL of recombinant hIFN-β for 48 h (Fig. 6A), an effect that was observed in a number of different cell lines in vitro, but only if their IFN signaling pathway was intact (Fig. 6B). Included in the panel of cell lines tested were 2fTGH human fibrosarcoma cells and two subpopulations of this parental line, U3A and U5A. These latter two cell lines are both defective in the type I IFN signaling pathway due to the loss of signal transducers and activators of transcription 1 and the Ifnar 2.2 component of the IFN-β receptor, respectively. The effect of continuous IFN-β treatment on angiopoietin-1 expression by tumors in vivo was then evaluated. Real-time PCR analysis, using human specific primers, of RNA extracted from U87 and NB-1691 tumors which showed restricted growth in mice treated with AAV IFN-β showed increased expression of angiopoietin-1 (Fig. 6C). This was confirmed at the protein level by Western blot analysis (Fig. 6C). To evaluate the time course of these changes effected by IFN-β, additional tumor-bearing mice were treated with AAV IFN-β. Tumors were harvested 0, 4, 7, 10, and 14 days later with relative levels of angiopoietin-1 induction and SMA-positive cell recruitment being assessed. Angiopoietin-1 induction was noted by 4 days after AAV IFN-β administration and reached a peak of >3-fold induction by 7 days. Increased SMA staining in IFN-β–treated tumors was noted by day 7 and had increased 4-fold over baseline by day 14 (Fig. 6D).

To confirm the central role of tumor-elaborated angiopoietin-1 in promoting IFN-β–mediated vascular maturation, established retroperitoneal NB-1691 xenografts were treated with both IFN-β and a truncated, soluble form of the Tie2 receptor that acts as a competitive inhibitor of angiopoietin-1 (28). Continuous systemic delivery of both IFN-β and Tie2 was achieved using separate liver-targeted AAV vectors (5 × 10¹⁰ vector particles of AAV IFN-β and AAV Tie2 per mouse given via tail vein). This produced stable levels of hIFN-β and Tie2 of 36.8 ± 5.6 and 6.2 ± 0.8 ng/mL, respectively, 1 week following vector injection. We found that maturation of the intratumoral vasculature, effected by systemic delivery of IFN-β, was abrogated by Tie2, as judged by a decrease in the intratumoral perivascular cell content in tumors treated with both IFN-β and Tie2 (Fig. 7A). In addition, the functional consequence of IFN-β–mediated remodeling of the intratumoral vasculature, improved drug delivery, was also partly reversed by Tie2 with topotecan tumor penetration in tumors treated with both IFN-β and Tie2, being ∼80% of that achieved in tumors treated with IFN-β alone (tumor topotecan concentration: control, 15.5 ± 0.3 ng/mL; IFN-β, 26.2 ± 0.5 ng/mL; IFN-β + Tie2, 23.4 ± 0.5 ng/mL; Fig. 7B).

Because the truncated soluble Tie2 used to block angiopoietin-1 signaling can also bind other angiopoietins and act as a competitive inhibitor to them, we also determined whether there were any changes in their expression in response to IFN-β. Although there was not a significant increase in angiopoietin-2 expression in treated U87 tumors, we did find, similar to angiopoietin-1, that expression of angiopoietin-2 was also increased in the NB-1691 tumors treated with IFN-β (Fig. 8). Angiopoietin-3 expression was unchanged in both tumor types following IFN-β treatment, and angiopoietin-4 expression was undetectable in both control and IFN-β–treated NB-1691 and U87 tumors.

Type I IFNs were the first endogenous cytokines tested in the biological therapy of cancer and have since become the most widely used (1). Their popularity stems from their potent antitumor activity in preclinical models and diverse mechanisms of action. In this study, we have identified a previously unknown antitumor mechanism of action for IFN-β, mediated through an apparently indirect effect on the tumor vasculature. When delivered in a continuous manner using a gene therapy–mediated approach, IFN-β promoted maturation and improvement in the functional efficiency of the vasculature of established tumors. Because angiogenesis is a dynamic process involving sequential stabilization and regression of newly forming tumor vessels during expansion of the vasculature (29), we speculate that the growth restriction of established tumors observed with this approach may have occurred because the stabilized tumor vasculature, reinforced with investing mural cells, could not be remodeled, thereby restricting further new vessel formation needed to support the increasing needs of a growing tumor. Although there may also have been some direct tumor cell cytotoxic activity of IFN-β in our tumor model, there were clearly profound effects on vessel morphology and physiology as a consequence of continuous delivery of IFN-β. The method of IFN-β dosing, using AAV vectors to attain continuous delivery of IFN-β, is likely responsible for the different antiangiogenic mechanism of action observed in our study, as we were unable to achieve the same antitumor efficacy or vascular remodeling with frequent administration of recombinant IFN-β protein (30).

In an attempt to explain the apparent paradox of synergy between agents that target the tumor vasculature and those that require intact blood vessels for efficient delivery, Jain (31, 32) hypothesized that antiangiogenic agents, rather than immediately destroying the tumor vasculature, can actually improve its function, at least transiently, thereby increasing tumor perfusion, and, consequently, delivery of cytotoxic agents. Most antiangiogenic agents that effect vascular remodeling seem to do so by inhibiting the vascular endothelial growth factor signaling pathway, often pruning inefficient neovessels while sparing the mature, more efficient vessels (33). In our model, however, vascular endothelial growth factor expression was not decreased in either the subcutaneous U87 or retroperitoneal NB-1691 tumors treated with continuous IFN-β (data not shown). However, both tumor types had significantly increased perivascular cell coverage of the tumor vessels, which was nearly absent in control treated tumors, whereas the mean vessel density either remained the same (U87) or actually increased (NB-1691). Thus, IFN-β treatment seemed to actively effect maturation of the vasculature rather than having achieved “normalization” through the pruning of immature vessels. In addition, with continued delivery of IFN-β, this normalization persisted for an extended length of time, during which improved delivery of conventional chemotherapy could be achieved. It is uncertain what direct inhibitory activity continuous delivery of IFN-β would have on activated endothelial cells in angiogenic tumor vessels and how this would affect the process of tumor vessel arterialization when the IFN-β used is host species compatible, however.

Molecular analysis revealed increased expression of angiopoietin-1 by the tumor cells in response to IFN-β. This effect was observed widely across a variety of solid human tumor-derived cell lines in vitro. Unlike the up-regulation of angiopoietin-1 expression in tumors treated with DC101, an anti–vascular endothelial growth factor receptor-2 monoclonal antibody used to achieve vascular endothelial growth factor signaling blockade and effect transient vascular remodeling (34), angiopoietin-1 induction on exposure of the cells to IFN-β in vitro suggests that this is a direct effect of IFN-β on the tumor cells. In vivo experiments showed that vessel maturation in response to IFN-β could be impaired by interfering with angiopoietin signaling using a soluble, truncated Tie2 receptor, further supporting a central role for angiopoietin-1 in the process of vascular normalization effected by IFN-β. These findings are consistent with previous studies that have suggested a role for angiopoietin-1 in vessel stabilization (27) and decreased vessel permeability (35) and reveal a novel method for effecting angiopoietin-1 up-regulation through continuous delivery of IFN-β. Although Tie2 blockade seemed to completely reverse perivascular cell recruitment to tumor vessels, this only partially abrogated the improvement in intratumoral drug delivery, suggesting that other factors, in addition to perivascular cell coverage of tumor vessels, may contribute to the functional improvement of these vessels in response to continuous IFN-β delivery. Interestingly, we found that continuous IFN-β delivery also augmented angiopoietin-2 expression in treated xenografts of one of the cell lines. Although angiopoietin-2 is reported to be an antagonist of angiopoietin-1 by preventing angiopoietin-1-mediated activation of the Tie2 receptor (25), several studies have shown that angiopoietin-2 can, in certain circumstances, actually activate Tie2 (36, 37). Because the soluble, truncated Tie2 receptor used in our studies binds both angiopoietin-1 and angiopoietin-2, we were not able to further define their relative contributions to the vascular maturation achieved with continuous IFN-β.

In summary, these results highlight the enhanced intratumoral delivery of cytotoxic drugs that can be achieved when given in combination with agents that effect vascular remodeling and maturation, a newly identified mechanism of action for IFN-β when delivered in a continuous manner. This effect on the vasculature seemed to be sustained, thereby facilitating improved drug delivery during a prolonged time period. This has significant implications for clinical cancer therapy in an effort to improve the antitumor activity of conventional cytotoxic agents, particularly given the broad responsiveness of different tumor histologies to IFN-β, with resultant up-regulation of angiopoietin-1 expression. Although the ultimate clinical potential of this gene therapy approach is uncertain, and probably awaits advances in vector targeting and regulation of transgene expression, it has very clear utility for studying the molecular and cellular factors involved in vessel remodeling and maturation mediated by continuous delivery of IFN-β.

The human glioblastoma (U87), hepatocellular carcinoma (HuH7), and Kaposi sarcoma (KS) cell lines and 293T cells (human embryonic kidney cells expressing SV40 large T antigen) were purchased from American Type Culture Collection. 2fTGH, U3A, and U5A fibrosarcoma cell lines were provided by G. Stark (Cleveland Clinic, Cleveland, OH). The human neuroblastoma cell line NB-1691 was provided by P. Houghton (St. Jude Children's Research Hospital, Memphis, TN) and subsequently modified to stably express luciferase, as described (18).

Tumor cells were plated at a density of ∼3 × 10⁵ per well. Recombinant hIFN-β (Avonex, Biogen, Inc.) was added to the culture medium daily. Cells were harvested after 48 h and RNA extracted for subsequent real-time PCR analysis. All experiments were done in triplicate.

Construction of the pAV2 hIFN-β, pAV2 FIX, and pAV2 Tie2 vector plasmids has previously been described (30). These vector plasmids each include the cytomegalovirus immediate-early enhancer, β-actin promoter, a chicken β-actin/rabbit β-globin composite intron, and a rabbit β-globin polyadenylation signal mediating the expression of the cDNA for hIFN-β, hFIX, or the truncated, soluble murine Tie2 receptor. The hIFN-β cDNA was purchased from InvivoGen; the hFIX cDNA was provided by G. Brownlee (University of Oxford, Oxford, United Kingdom); and the Tie2 cDNA was provided by P. Lin (Vanderbilt University Medical Center, Nashville, TN). Recombinant AAV vectors pseudotyped with serotype 8 capsid were generated by the method previously described using the pAAV8-2 plasmid provided by J. Wilson (University of Pennsylvania Medical Center, Philadelphia, PA; ref. 38). These AAV2/8 vectors were purified using ion exchange chromatography (39).

Subcutaneous (heterotopic) glioma xenografts were established in male C.B-17 severe combined immunodeficient mice (Jackson Laboratory, Bar Harbor, ME) by injection of 1.5 × 10⁶ U87 tumor cells in 150-μL PBS into the subcutaneous space of the right flank. Retroperitoneal (orthotopic) neuroblastoma xenografts were established by injection of the same number of NB-1691 cells in 150-μL PBS behind the left adrenal gland via a left subcostal incision during administration of 2% isoflurane. Measurements of the subcutaneous tumors were made with calipers and measurements of the retroperitoneal tumors were made by ultrasonography. With both modalities, measurements were done in two dimensions and volumes calculated as width² × length × 0.5. Bioluminescence imaging of NB-1691/luciferase tumor–bearing mice was done using an IVIS Imaging System 100 Series (Xenogen Corp.). Animals treated with topotecan (GlaxoSmithKline) received 2 g/kg via tail vein injection. 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.

Systemic levels of hIFN-β and murine Tie2 in mouse plasma were determined using commercially available immunoassays (ELISA; Biosource International and R&D Systems, respectively). In selected plasma samples, the biological activity of the expressed hIFN-β was analyzed using an antiviral assay in which protection against the cytopathic effect of vesicular stomatitis virus on human fibroblasts was determined (40).

Plasma levels of murine alanine aminotransferase were measured with the Vitros DT 60 chemistry system (Ortho-Clinical Diagnostics).

Real-time PCR. Assay-on-Demand gene expression primer and probe sets for human angiopoietin-1, angiopoietin-2, angiopoietin-3, angiopoietin-4, and glyceraldehyde-3-phosphate dehydrogenase (Hs00919202_m1, Hs01048042_m1, Hs00559786_m1, Hs00907078_m1, and Hs9999905_m1, respectively; Applied Biosystems, Foster City, CA) were used for real-time PCR done on an ABI 7900HT for 40 cycles (50° for 2 min, 95°C for 10 min, 95°C for 15 s, 60°C for 60 s). Relative gene expression was calculated with the ΔΔC_t method normalized against glyceraldehyde-3-phosphate dehydrogenase and using the untreated sample as calibrator.

Western Blot Analysis. Rabbit anti-human angiopoietin-1 and angiopoietin-2 antibodies (provided by Regeneron), used at a concentration of 0.45 μg/mL, were used for protein detection in Western blot analyses.

Tumor Immunohistochemistry. Formalin-fixed, paraffin-embedded tumor sections (4 μm thick) were stained with rat anti-mouse CD34 (RAM 34, PharMingen) and mouse anti-human SMA (clone 1A4, DAKO) antibodies (41). Sections were viewed and digitally photographed using an Olympus U-SPT light microscope with an attached charge-coupled device camera. Four images at ×100 were taken of each tumor section with care to avoid areas of necrosis. Images were saved as JPEG files for further processing in Adobe Photoshop (Adobe Systems, Inc.). Positive staining was quantified using NIH image analysis software ImageJ and is reported as the mean number of positive pixels per tumor section. Briefly, images were processed by removal of background hematoxylin staining followed by monochromization. Monochrome images (giving binary data: black areas of immunohistochemical stain and white nonstaining areas) were analyzed using ImageJ software, which calculated the absolute values for each channel, resulting in determination of percentage of immunopositivity per unit area. A threshold value of 75 grayscale units was applied to captured images and used to determine positivity. To assess intratumoral hypoxia, mice were injected i.p. with Hypoxyprobe-1 (Chemicon International) 90 min before sacrifice. Tumors were then harvested, fixed, and embedded. Tumor sections were stained using the Hypoxyprobe-1 Plus Kit (HP2-100, Chemicon). Hypoxia positive staining was also quantified using the ImageJ software, as described above. Because tumor hypoxia can be affected by tumor volume, size-matched tumors were used for comparisons. Immunohistochemical staining for PDGFRβ was accomplished with a goat polyclonal antibody directed against mouse PDGFRβ (R&D Systems) as previously described (42).

Confocal Microscopy. Tumor-bearing mice were injected via tail vein with a 2,000,000 MW FITC-labeled dextran (Molecular Probes), which was allowed to circulate for 3 min. The vasculature was then perfused with 4% paraformaldehyde fixative. Tumors were harvested, fixed overnight in the same fixative, and embedded in a 10% agarose gel. Sections (100 μm) were cut with a vibratome and incubated overnight with a Cy3-conjugated anti-SMA monoclonal antibody (clone 1A4, Sigma) diluted 1:1,000 in PBS containing 0.01% thimersol and 0.3% Triton X-100. Sections were then washed with sterile PBS and analyzed using a Leica TCS SP confocal microscope.

Intravital Microscopy. Mice bearing subcutaneous U87 tumors were anesthetized with ketamine/xylazine (0.1 mL s.c.) and the skin overlying the tumors was incised to expose the tumors. The superficial tumoral vasculature was visualized with an industrial scale microscope (model MM-40, Nikon USA) with a digital camera (Photometric CoolSnap FX, Roper Scientific) and fluorescent (100 W, mercury) light source. Images were acquired at ×40 magnification and analyzed using MetaMorph software (Universal Imaging Co.; ref. 43).

Evan's Blue Dye Assay. One hundred microliters of 2% Evan's blue dye (MP Biomedicals) were administered via tail vein of tumor-bearing mice and allowed to circulate for 20 min. To remove remaining intravascular dye, mice were perfused with 10 mL of saline through the left ventricle with a right atrial vent. Because variability in the pressure of perfusion could alter the results, the same perfusion setup, volume of perfusate, and time (30 s) for the perfusion of each animal were kept constant. Excised tumors were cut into blocks weighing 0.15 g, which were then placed in 1 mL of formamide (Fisher Scientific) for 72 h for Evan's blue extraction. These tumor specimens were then removed and resultant extract was centrifuged. Levels of Evan's blue were quantified using a spectrophotometer at a wavelength of 620 nm. A standard curve was generated using 0.15-g samples of tumor tissue from animals that had not been given Evan's blue dye (but were perfused with saline), which were placed in formamide together with a serial dilutions of the dye. All samples were run in duplicate and compared with those of standards. Results are expressed as micrograms per milligram of tumor calculated by averaging the extracted values for the sections from each tumor.

Tumor Interstitial Fluid Pressure. Tumor interstitial fluid pressure was determined in subcutaneous tumors using a needle pressure technique. A 23-gauge hollow bore needle with an additional side hole cut 5 mm from the tip was connected to saline filled tubing that was then connected via a fluid filled dome diaphragm (Memscap AS) to a MLT844 physiologic pressure transducer (ADInstruments, Inc.). The transducer is connected to a ML110 bridge amplifier (ADInstruments) and data are imported into PowerLab software (ADInstruments). Before and in between recordings of individual tumor pressures, the system was calibrated using a graduated water column. Data are reported in millimeters of mercury (1 cm H₂O = 1.36 mm Hg). Because the interstitial fluid pressure reading can vary depending on where in a tumor the needle is inserted, we tried to ensure that the location of the needle was consistent. The width of each tumor was measured and the needle was inserted into the tumor to a depth equal to one third the width, ensuring placement well within the tumor but not directly in the potentially necrotic tumor center.

Contrast-Enhanced Ultrasonography. Optison ultrasound contrast agent (Amersham Health, Inc.) was used to carry out contrast-enhanced ultrasonography, as described (20). Optison is a suspension of human serum albumin microspheres encapsulating octafluoropropane gas. Mean microsphere size ranges from 2 to 4.5 μm and particles remain in the intravascular space. Briefly, a region of interest within the ultrasound image was drawn to encompass the entire tumor, which was then evaluated for precontrast baseline signal intensity, change in signal intensity from baseline to initial peak (in decibels), and rate of signal intensity increase from baseline to initial peak (in decibels per second).

One hour after bolus intravenous topotecan administration, whole blood (in EDTA) and tumor tissue were collected. Snap-frozen tumors were homogenized in PBS (100 μL/mg of tissue) and centrifuged to remove particulates. Cold methanol (4:1) was added to the supernatant and the plasma collected from the whole blood to extract the topotecan. Samples were centrifuged and the methanolic supernatant was stored at −80°C until analysis. The extracted tumor and plasma samples were converted to total lactone by adding five parts methanolic supernatant to one part 20% phosphoric acid. The methanolic supernatant from tumor tissue or plasma was analyzed for topotecan by an isocratic high-performance liquid chromatography with fluorescence detection (RF10AXL; Shimadzu); the excitation and emission wavelengths were 370 and 530 nm, respectively (44).

Results are reported as mean ± SE. The SigmaPlot program (SPSS, Inc.) was used to analyze and graphically present the data. Unpaired Student's t test was used to analyze statistical differences in the results.

We thank Junfang Zhou for technical assistance, Dorothy Bush and Colleen Anderson for their assistance with immunohistochemistry, Stacey Glass for her assistance with ultrasonography, Nikolaus Hagedorn and Charles Fraga for their assistance with topotecan measurements, Dr. Gopal Murti for his assistance with confocal microscopy, and Dr. John Gray and the staff of the Vector Core Facility at St. Jude Children's Research Hospital for their assistance in generating pseudotyped AAV vectors particles required for this study.