Vascular Normalization by Vascular Endothelial Growth Factor Receptor 2 Blockade Induces a Pressure Gradient Across the Vasculature and Improves Drug Penetration in Tumors

The present treatment of solid tumors is plagued by two problems: physiological barriers impair the delivery of therapeutic agents in optimal quantities, and genetic and epigenetic mechanisms contribute to drug resistance. Antiangiogenic therapy has the potential to overcome or circumvent these problems. A recent Phase III clinical trial (1) has provided the first clinical proof of this hypothesis, the combination of Avastin, a monoclonal antibody against vascular endothelial growth factor (VEGF), with conventional chemotherapy increased disease-free survival by 5 months in colorectal cancer patients. In addition to the direct antivascular effects that lead to the death of cancer cells, we proposed that anti-VEGF antibody improves the delivery of cytotoxic agents to tumors and, thus, increases the effectiveness of combination therapy (2). We have recently shown that an anti-VEGF antibody can lower interstitial fluid pressure (IFP) in transplanted tumors in mice (3) and in rectal carcinomas in patients (4). However, the mechanisms of IFP reduction and how this reduction affects drug delivery are not known. In normal tissues, the excess fluid filtered from blood vessels is drained by lymphatic vessels to maintain the IFP close to zero (mm Hg). In tumors, IFP homeostasis is perturbed because of impaired lymphatic function (5, 6) and abnormalities in vascular structure and function (7). Because of the high vascular permeability and the impaired lymphatic drainage, the oncotic and hydrostatic pressures in the microvascular and interstitial spaces are at equilibrium in tumors (8, 9). To unravel the mechanism of IFP reduction induced by blocking VEGF signaling, we examined the effect of DC101 on the determinants of interstitial hypertension, including changes in the morphology, wall structure, and function of tumor vasculature during the course of treatment.

The murine mammary carcinoma MCaIV, human small cell lung carcinoma 54A, human glioblastoma multiforme U87, and human colon adenocarcinoma LS174T were grown in nude or severe combined immunodeficient (SCID) mice. MCaIV, 54A, and U87 were transplanted s.c. MCaIV and LS174T tumors were also transplanted in dorsal skinfold chambers in SCID mice for intravital microscopy and for microvascular pressure (MVP) and IFP measurements. Animals were anesthetized with ketamine/xylazine (100/10 mg/kg, i.m.) unless otherwise noted.

DC101 (ImClone Systems Inc., New York, NY) was administered i.p. at a dose of 40 mg/kg, previously determined to have antitumor effects in vivo (10). The control group received 40 mg/kg, i.p of nonspecific rat IgG1 antibody (ImClone). All of the experiments were done 3 days after the injection of DC101 (IgG for the control), unless otherwise noted.

To identify the functional lymphatic vessels in tumors, ferritin microlymphangiography and lymphatic vessel endothelial hyaluronan receptor-1 (LYVE-1) immunostaining were performed according to published methods (6).

To visualize functional blood vessels in MCaIV or LS174T in the dorsal chamber of SCID mice in vivo, fluorescein-labeled dextran (M_r 2,000,000; Sigma) was injected i.v. (11). Vascular length per unit area and vessel mean diameter were obtained by averaging measurements over six random regions per mouse. The effective vascular permeability was measured using cyanine-5-labeled BSA as described previously (12). For multiphoton microscopy images, maximum intensity projections of stacked images were done offline using Confocal Assistant.

MCaIV-tumor-bearing mice (n = 6 in control group; n = 5 in DC101-treated group) were injected with FITC-CD31 1 h before perfusing fixative. The mice were perfused and fixed with 4% paraformaldehyde, and 100-μm-thick sections were prepared (13). Collagen IV (Chemicon) was stained overnight; and tetramethyl rhodamine isothiocyanate (TRITC) goat antirabbit antibody was used as a secondary antibody. Three images per section were taken by confocal microscopy. Total vessel length and the length of segments covered by collagen IV were traced in three random regions (each region was 700 μm by 700 μm) for each 100-μm-thick section. On average, ∼50 vessel segments were measured per mouse.

Similarly, MCaIV-tumor-bearing mice (n = 4) were given injections of FITC-CD31 and were perfused and fixed with 4% paraformaldehyde, and 100-μm-thick sections were prepared. Anti-α-smooth muscle actin (αSMA; Cy-3-conjugated mouse monoclonal, clone 1A4; Sigma) antibody was used to identify αSMA-positive perivascular cells. The tissue sections were examined using a confocal microscope. Quantification was performed in a manner similar to collagen IV staining.

IFP was measured with the wick-in-needle technique (14) prior to and during DC101/control antibody treatment for 54A and U87. DC101/control antibodies were injected once every 3 days. For each time point, IFP was measured in at least two different tumor regions.

MCaIV tumors were implanted in the dorsal skinfold chamber of SCID mice as described previously (15). The dorsal skinfold chamber preparation provides a stable and easy access for IFP and MVP measurements with the micropipette technique (8, 16). For each mouse, two to seven IFP and MVP measurements were obtained. Under stereomicroscopic guidance, micropipettes with an opening of ∼2.5 μm were inserted inside blood vessels to measure the MVP (confirmed by Lissamine Green B injection). The IFPs were generally measured at 0.5–1 mm from the tumor surface.

The chronic wick technique was used to collect tumor interstitial fluid (9). A long wick (∼5 cm) was transplanted together with the MCaIV tumors and was collected postmortem 2 weeks later; interstitial fluid was obtained by centrifugation. A membrane colloid osmometer with ultrafiltration membranes (Amicon PM10; Millipore Corp., Bedford, MA) was used to measure oncotic pressure.

TRITC-BSA (0.1 ml; Molecular Probes) was injected i.v. 1 h before perfusion fixation in tumor-bearing mice. Biotinylated lectin was also injected 5 min before fixation. Three 10-μm-thick frozen sections, 50 μm apart, were prepared from each tumor block. Images were taken using two-photon microscopy. Perfused biotinylated lectin was used to identify functional vessels, and extravasation pattern of TRITC-BSA was analyzed using a program written in ImageJ. Ten concentric rings of 3.25-μm thickness were drawn starting at the vessel wall. Average intensity was calculated within each ring, and the intensity profile was fitted to an exponential function (I = Ae^−x/L + C, where I = pixel intensity, x = distance from vessels, and L = characteristic penetration length). The characteristic penetration length was compared between the two treatment groups.

Data are presented as mean ± SE. Significant differences between groups were determined with a Student’s t test (JMP Program). P < 0.05 was considered statistically significant.

We first confirmed that similar to a VEGF blocking antibody (3), DC101 decreased IFP in both human small cell lung carcinoma 54A (Fig. 1,A), and human glioblastoma multiforme U87 xenografts (Fig. 1,B). At the same time, we did not find any evidence of functional lymphatic vessels in the tumors of either treated or control animals (Fig. 1 C). Thus, the drop in IFP could not be attributed to modifications in lymphatic function.

We then examined the effect of DC101 on the morphology of blood vessels in the murine mammary carcinoma MCaIV growing in dorsal skinfold chambers. Three days after a single injection of DC101, the vascular density and vessel diameter decreased significantly (Table 1; Fig. 2, A and B). To obtain further insight into the dynamic changes occurring in the tumor vasculature, we continuously monitored the same region of the tumor over 6 consecutive days using two-photon microscopy. This technique allowed us to image up to 200 μm deep with 1-μm resolution (11). In contrast to skeletal muscle, which has an organized vasculature with a relatively smooth vascular wall and uniform diameter (Fig. 2,C), untreated MCaIV tumors had tortuous vessels with abrupt changes in vessel diameter (arrowheads in Fig. 2,D). On the basis of the dynamic high resolution images, the surprising finding was that, 2–3 days after DC101 treatment, many of these vessels also became less tortuous besides getting smaller in diameter (arrow in Fig. 2,E). By day 5, in some regions, some of these vessels regressed completely. Furthermore, the vessels at the tumor–host interface also became less tortuous and assumed a relatively normal morphology (Fig. 2,F). Similar striking vascular changes were also observed in the human colon adenocarcinoma LS174T (Fig. 2, G and H). Thus, DC101 “normalizes” the architecture of the vascular network before complete regression.

In addition to architectural abnormalities, the tumor vasculature is characterized by a paucity of, or abnormalities in, mural cells and basement membrane (7, 13, 17). To test whether DC101 also normalized the wall structure, we perfused the tumors with FITC-CD31 and stained frozen sections for collagen IV and αSMA (which label basement membrane and mural cells, respectively). Unlike the vessels in normal muscle, the collagen IV staining around some vessels was below the detection limit in MCaIV tumors (Fig. 3,A). Consistent with normalization, more vessels exhibited collagen IV staining after DC101 treatment (Fig. 3, B and C). DC101 also significantly increased the fractional coverage of tumor blood vessels by αSMA-positive cells (Fig. 3 D). Approximately 25% of the vessels in untreated tumors had little or no perivascular cell coverage, but this fraction dropped to ∼8% after DC101 treatment (P < 0.005). Thus, DC101 induces structural normalization of the tumor vasculature by homogenizing the vessel size, reducing vessel tortuosity, and increasing the fractional coverage of the vascular endothelium by perivascular cells and basement membrane.

To determine whether structural normalization of tumor vessels translates into functional normalization, we measured the functional parameters that govern the transport of molecules across the vessel wall. These include vascular permeability, plasma and interstitial oncotic pressures, MVP, and IFP (18, 19). Three days after the injection of DC101 the vascular permeability of albumin in MCaIV, a profoundly leaky tumor, was decreased by 51% (Table 1). In general, leaky vessels in tumors lead to an increase in the interstitial oncotic pressure (9), which becomes approximately equal to the plasma oncotic pressure. We found that in DC101-treated tumors the interstitial oncotic pressure was significantly lower than in the control group (Fig. 3 E), whereas the plasma oncotic pressure did not change. This decrease in interstitial oncotic pressure is consistent with the decrease in vascular permeability to macromolecules induced by DC101. The enhanced oncotic pressure gradient across the vasculature is another hallmark of tumor vasculature normalization.

In solid tumors, MVP is approximately equal to IFP, leading to nearly zero pressure difference across the vessel wall (8). As a matter of fact, these two pressures are so closely coupled that changing the vascular pressure leads to similar changes in the IFP within seconds, until they both become equal again (20). Thus the DC101-induced decrease in IFP might be accompanied by a similar decrease in MVP. To test this, we measured MVP in control and DC101-treated tumors by directly inserting micropipettes in MCaIV tumor vessels (16). We found that MVP was unaffected by DC101 treatment (Fig. 3 F). Thus, DC101 creates a sustained hydrostatic pressure gradient across the vasculature.

Movement of molecules across vessel walls occurs by diffusion and convection. Diffusion is governed by concentration gradients across the vessel wall, whereas convection is governed by pressure gradients. To test whether the pressure gradient across the vascular wall improves the penetration of large molecules, we injected fluorescently-labeled BSA i.v. into MCaIV-tumor-bearing mice and observed the pattern of BSA extravasation. We also identified functional blood vessels using biotinylated lectin. Quantitative analysis confirmed that DC101 produced a significantly deeper penetration of BSA molecules into the tumor (Fig. 4, A–E).

The data presented here provide compelling evidence in support of the normalization hypothesis: that judicious application of antiangiogenic treatment can normalize the tumor vasculature by pruning the immature vessels and fortifying the remaining ones (2). We show that the normalized tumor vasculature is less tortuous and the vessels are more uniformly covered by pericytes and basement membrane (Fig. 4 F). This result implies that vessels with less mural cell coverage are more vulnerable to DC101-induced regression, which is consistent with previous findings that tumor vessels without mural cells tend to regress after VEGF withdrawal (21). Our functional data demonstrate for the first time that a drop in vascular permeability induces transvascular gradients in oncotic and hydrostatic pressure in tumors. The induced hydrostatic pressure gradient improves the penetration of large molecules into tumors.

In tumors, the abnormally high leakiness of the vasculature hinders drug delivery by inducing blood flow stasis (22, 23), whereas normalized vessels are less leaky to macromolecules. However, the permeability of the normalized vasculature is still significantly higher than that of normal tissues (12, 18). The small drop in oncotic pressure from 19 to 17 mm Hg induced by DC101 is also indirect evidence that the vascular permeability of DC101-treated MCaIV tumors is still elevated. For example, in s.c. tissue, the oncotic pressure is ∼8 mm Hg (9). The change in oncotic pressure associated with DC101 leads to a minor increase in the hydrostatic pressure gradient, which is sufficient to increase the transvascular convection of a macromolecule and is most likely responsible for the deeper penetration of albumin.

Fluid movement across the vascular wall can be described by Starling’s equation (18). Starling’s equation states that the rate of fluid movement across a unit area of vascular wall (J_v/A) is proportional to the net pressure difference across the vessel wall:

where L_p (the hydraulic conductivity of the vessel wall) is the proportionality coefficient and A is the vascular wall area. As shown in the equation, the hydrostatic pressure gradient (MVP − IFP) is balanced by the oncotic pressure gradient [the difference between π_p (plasma oncotic pressure) and π_i (interstitial oncotic pressure)]. The reflection coefficient, σ, determines the effectiveness of the oncotic pressure gradient across the vascular wall. The value of σ for albumin varies between 0 for the liver (with a high vascular permeability) and 1 for the impermeable brain vessels; lung has a σ of 0.5 (24, 25). To our knowledge, the reflection coefficient of the tumor vasculature has not been measured. Because the tumor vasculature has relatively large pores (26) and the oncotic pressure in the interstitium is similar to the oncotic pressure of plasma, the reflection coefficient is most likely close to 0. Based on the theory for hindered transport of rigid solutes, the calculated reflection coefficient for BSA is 0.0002 (σ = [1 − [1 − radius of BSA (3.5 nm)/vessel pore size (500 nm)]²]²) (26). Thus, the increase in oncotic pressure gradient induced by DC101 has minimal effects on the transvascular flow.

We have previously shown that angiotensin II can increase the systemic blood pressure in mice and can create a pressure gradient across the vessel wall in tumors (20). Unfortunately, interstitial pressure catches up with vascular pressure, and the pressure gradient across the vessel wall dissipates in less than 1 min. Nevertheless, even these short-lived gradients can increase the delivery of specific antibodies in tumors (20). Thus, the sustained hydrostatic pressure gradient across the tumor vasculature, induced by DC101, can increase the transvascular convection of large molecules, despite a drop in vascular permeability to macromolecules.

The original rationale for combining antiangiogenic and cytotoxic therapies was to target two distinct cell populations within solid tumors: cancer cells and endothelial cells. When endothelial cells are targeted, blood vessels should be destroyed, and, thus, the delivery of therapeutics is compromised. However, several preclinical studies clearly demonstrate that the delivery of therapeutics is not compromised by antiangiogenic agents (27), and it even increases in some cases (28). The vessel normalization and the restoration of pressure gradients induced by VEGF blockade may explain the increased uptake of cytotoxic agents in tumors. For a cytotoxic agent to be effective, it must reach all cancer cells and in effective quantities. Antiangiogenic therapy, as shown by this study, might facilitate a more uniform delivery of therapeutic agents to cancer cells, including those that are farther from the vessels. This mechanism might contribute to the potentiation of conventional therapies by antiangiogenic agents (1, 10, 28, 29, 30, 31).

To dissect molecular mechanisms involved in the normalization process, we collected total RNA and protein from MCaIV tumors 3 days after the completion of treatment with DC101 or control IgG. cDNA gene chips were used for screening (Angiogenic Gene Array, SuperArray, Bethesda, Maryland). Based on the gene array data, multiple genes were found to be differentially expressed after DC101 treatment (e.g. 2.2 fold downregulation of Ang2). Real time PCR data (LUX system, InVitrogen, Carlsbad, California) only confirmed the significant difference (P < 0.005) in Ang2 expression between control (n = 8, 1.29 ± 0.22) and DC101-treated group (n = 8, 0.29 ± 0.06). Furthermore, by Western analysis there was a marked decrease in the level of Ang2 protein after DC101 treatment. Immunostaining revealed Ang2 expression in both tumor and stromal cells. These data suggest that VEGFR2 blockade induces a decrease in Ang2 expression in tumors. Ang2 has been shown to destabilize blood vessel (7), thus the reduction in Ang2 expression may contribute to a more stable vasculature as described above. Further studies are needed to understand the crosstalk between the VEGF and angiopoietin pathways.

We thank Eugene M. Renkin (University of California Davis, Davis, CA) for his generous gift of the nylon wick material for interstitial fluid collection; T. Padera for lymphatic vessel endothelial hyaluronan receptor-1 (LYVE-1) staining; R. Delgiacco, P. Huang, J. Kahn, S. Roberge, and C. Smith for the outstanding technical support; S. Kwei for image analysis; M. Booth, E. Brown, W. Deen, D. Duda, D. Fukumura, R. Jones, L. Munn, J. Samson, M. Stohrer, E. di Tomaso, and J. Yuen for helpful comments.