Functional and molecular mapping of uncoupling between vascular permeability and loss of vascular maturation in ovarian carcinoma xenografts: The role of stroma cells in tumor angiogenesis



Maintaining homogeneous perfusion in tissues undergoing remodeling and vascular expansion requires tight orchestration of the signals leading to endothelial sprouting and subsequent recruitment of perivascular contractile cells and vascular maturation. This regulation, however, is frequently disrupted in tumors. We previously demonstrated the role of tumor-associated myofibroblasts in vascularization and exit from dormancy of human ovarian carcinoma xenografts in nude mice. The aim of this work was to determine the contribution of stroma- and tumor cell-derived angiogenic growth factors to the heterogeneity of vascular permeability and maturation in MLS human ovarian carcinoma tumors. We show by RT-PCR and by in situ hybridization that VEGF was expressed by the tumor cells, while angiopoietin-1 and -2 were expressed only by the infiltrating host stroma cells. Vascular maturation was detected in vivo by vasoreactivity to hypercapnia, measured by BOLD contrast MRI and validated by immunostaining of histologic sections to α-smooth muscle actin. Vascular permeability was measured in vivo by dynamic contrast-enhanced MRI using albumin-based contrast material and validated in histologic sections by fluorescent staining of the biotinylated contrast material. MRI as well as histologic correlation maps between vascular maturation and vascular permeability revealed a wide range of vascular phenotypes, in which the distribution of vascular maturation and vasoreactivity did not overlap spatially with reduced permeability. The large heterogeneity in the degree of vascular maturation and permeability is consistent with the differential expression pattern of VEGF and angiopoietins during tumor angiogenesis. © 2005 Wiley-Liss, Inc.

Tumor angiogenesis is characterized by extensive structural and functional heterogeneity. Temporal changes include expansion and regression of blood vessels. Regression upon VEGF withdrawal was described previously for immature tumor vessels.1, 2 Periendothelial α-SMA-positive cells (pericytes and smooth muscle cells) were suggested to protect neovasculature selectively from obliteration upon VEGF withdrawal.3, 4 Nevertheless, not all α-SMA-positive cells are functional in maintaining vessel integrity. It was recently demonstrated in 3 tumor types that the majority of blood vessels were coated with pericytes, which were either desmin- or α-SMA-positive. In contrast with normal tissues, these pericytes were loosely attached and displayed multiple abnormalities.5 The position of pericytes in the leading edge of sprouting capillaries suggested that these cells might have a role in sprout growth or retraction.

The VEGF family members (VEGF A–D and PlGF) interact differentially with endothelial-specific receptors (VEGFR 1–3) to regulate vascular permeability, capillary sprouting and endothelial cell survival.6 In addition to the VEGFR family, angiogenesis is regulated also by TIE2, a cell surface receptor expressed specifically by endothelial cells.6 TIE2 ligands, Ang-1 and Ang-2, were suggested to collaborate with VEGF in regulation of vascular remodeling, permeability and recruitment of periendothelial cells.6, 7

VEGF-A is also known as vascular permeability factor (VPF).8 Using DCE-MRI, we have shown that acute intradermal administration of VEGF increases permeability to a macromolecular contrast material biotin-BSA-Gd-DTPA in normal skin.9 In C6-pTET-VEGF tumors, with tetracycline switchable overexpression of VEGF, overexpression of VEGF led to extravasation of the MR contrast material. This vascular hyperpermeability was significantly suppressed after VEGF withdrawal by administration of tetracycline to the drinking water.10 The degree of vascular maturation in this C6-pTET-VEGF165 tumor model was mapped by BOLD contrast MRI using vasoreactivity in response to hypercapnia as a biomarker for vascular maturation.4 Withdrawal of VEGF resulted in selective loss of functionality in immature vessels.4 Thus, MRI could be used to image noninvasively 2 distinct features of the neovasculature, namely, hyperpermeability and reduced vasoreactivity.

A number of studies attempted to correlate the vascular responses detected by BOLD contrast and by DCE-MRI, usually using hyperoxic hypercapnia (carbogen) for BOLD imaging and low-molecular-weight Gd-DTPA for DCE-MRI.11, 12, 13 These studies reported overlapping sensitivity to blood volume but otherwise dissimilar patterns of activation. In order to attain better the biologic understanding of the 2 imaging approaches, we used here normoxic hypercapnia so as to allow higher sensitivity to vasoreactivity, while a macromolecular contrast material that is also detectable by histology was used for increased specificity to vascular permeability and reduced effects of flow.

In other models, VEGF-induced vascular permeability could be blocked by overexpression of Ang-1 without altering vessel morphology.14, 15 Vascular defects in Ang-1 knockout mice16 and mice overexpressing Ang-1 in the skin17 demonstrated that Ang-1 plays a critical role in vascular wall stabilization through recruitment of supporting cells.6 Ang-2 was initially discovered as the natural antagonist of Ang-1 and was suggested to induce destabilization of mature vessels. The phenotype of mice overexpressing Ang-2 was similar to that of the Ang-1 knockout mice.18 Ang-2 was also shown to act as an agonist of Ang-1 in lymphatic development, while in the hyaloid system development, Ang-2 antagonizes Ang-1 activity. Thus, mice lacking Ang-2 were defective in regression of the hyaloid vascular system, indicating that Ang-2 is essential for vascular remodeling.19

Angiopoietins expression was observed in many tumors, including ovarian cancer,20 non small cell lung cancer,21 breast cancer22 and astrocytoma.23 In a model of human squamous cell carcinoma (SCC) xenografts, low levels of Ang-1 but not Ang-2 were detected in the skin. Ang-2 was upregulated by the tumor and was found to be expressed by endothelial cells. Overexpression of Ang-2 compensated for the Ang-1 inhibitory effect on tumor growth.24 In implanted C6 glioma spheroids, expression of Ang-2 was shown to be restricted to the host endothelium and was associated with vascular remodeling during tumor growth.25 Ang-2 and TIE2 expression by endothelial cells during vessel cooption and regression suggested that Ang-2 acts in an autocrine manner.26

Overexpression of Ang-1 inhibits tumor growth presumably through vascular stabilization and preventing the initiation of vascular sprouting.24, 27 Overexpression of Ang-1 was also reported to inhibit growth of MCF7 breast tumor xenografts in a dose-dependent manner. In that system, overexpression of Ang-1 stimulated infiltration of α-SMA-positive cells into the tumor and augmented microvessel coverage by pericytes.28

The role of stroma cells in tumor angiogenesis in general and in ovarian carcinoma in particular was evaluated, showing specific contribution of stroma cells to the expression of angiogenic growth factors29 as well as to the response to therapy.30 The role of stroma cells was demonstrated to depend on PDGF expression, allowing vascularization of VEGF-deficient tumors.31 Infiltration of endothelial cells into ovarian carcinoma tumors was dependent on the presence of myofibroblasts.32 In ovarian carcinoma, we have shown that expression of VEGF is induced by hypooxia as well as by hormonal stimulation.33 We have recently reported that dormancy of MLS human ovarian carcinoma spheroids implanted in nude mice was associated with extensive vascular remodeling.34 Exit from dormancy correlated with infiltration of α-SMA-positive stroma myofibroblasts, forming an extensive stroma network.35

The aim of this study was to explore the relationship between tumor vessel architecture, permeability and the degree of maturation in the context of VEGF and angiopoietins expression in growing MLS human epithelial ovarian carcinoma xenografts. We show here that while VEGF was expressed by tumor cells both in vitro and in vivo, expression of Ang-1 and Ang-2 was restricted to the host infiltrating stroma cells. The impact of such expression pattern would be that endothelial cells residing in different tumor regions will be exposed to different levels of VEGF, Ang-1 and Ang-2 (Fig. 1a). This is expected to lead to a large vascular heterogeneity and uncoupling between reduced vascular maturation, which is mainly driven by the ratio of Ang-2 to Ang-1, and increased permeability driven by VEGF, as was indeed detected here by MRI, fluorescence microscopy and histology.

Figure 1.

Differential expression of VEGF and angiopoietin can lead to heterogeneous pattern of vascular maturation and permeability. (a) Since VEGF is secreted by the tumor cells but Ang-1 and Ang-2 are secreted by infiltrating stroma cells, vessels in different regions would be exposed to different levels of these factors. In regions of high VEGF and low Ang-1/Ang-2 ratio, the vessels will be immature and hyperpermeable (green zone). High VEGF and high Ang-1/Ang-2 ratio will result in mature vessels and the permeability will be determined by the ratio of VEGF/Ang-1 (cyan; blue-green). Low VEGF and high Ang-1/Ang-2 ratio will lead to highly mature vessel, which shows low plasma leakage (blue zone). High Ang-2 in the absence of Ang-1 and VEGF will result in vessel regression (black). (b) Multimodality analysis of vascular maturation and permeability. Noninvasive BOLD contrast and DCE-MRI of vasoreactivity (indicative of maturation) and permeability to macromolecular contrast material (biotin-BSA-Gd-DTPA) were followed by histologic analysis of vascular maturation (recruitment of α-smooth muscle actin expressing cells) and vascular permeability (extravasation of the MR contrast material), and in situ hybridization analysis of the expression of VEGF and Ang-1.


α-SMA, α-smooth muscle actin; Ang-1, angiopoietin-1; Ang-2, angiopoietin-2; APS, apparent permeability surface area product; BOLD, blood oxygen level-dependent; DCE, dynamic contrast enhancement; MRI, magnetic resonance imaging; VEGF, vascular endothelial growth factor.

Material and methods

Cells and spheroid culture

MLS, a human epithelial ovarian carcinoma cell line, was cultured in α-MEM supplemented with 10% FCS and antibiotics (50 units/ml penicillin, 50 μg/ml streptomycin and 125 μg/ml fungizone; Biolab Ltd., Jerusalem, Israel). MLS spheroids were initiated by spontaneous aggregation of cells in suspension followed by stirred culture as described previously.33, 34 Spheroids of 1 mm in diameter were selected for further analysis.

Animal protocols

All animal investigations were conducted in accordance with the Guiding Principles for the Care and Use of Research Animals of the Weizmann Institute of Science. MLS tumors were generated by subcutaneous inoculation of 0.5–1 × 106 cells in the lower back of female CD1 nude mice (6- to 10-week-old; body weight, 28–30 g). For MRI studies, mice were anesthetized with 75 mg/kg ketamine and 3 mg/kg xylazine (i.p.), followed by subcutaneous addition of about 50% the initial dose in order to prolong the anesthesia. Tumor diameter did not exceed 7 mm.

MRI experiments

MRI scans included BOLD contrast detection of the response to hypercapnia followed by DCE-MRI using biotin-BSA-Gd-DTPA [see experimental timeline in Figure 1(b); n = 5]. Vascular maturation was determined from changes in BOLD contrast MRI in response to hypercapnia (inhalation of air enriched with 5% CO2). In this type of analysis, mature vessels will react to increased CO2 and therefore will affect the MR signal, yielding a map of the spatial distribution of mature vessels in the tumor.4, 34, 36 Vascular permeability maps (APS9, 10, 37) were generated from the slope of the MR signal enhancement after intravenous administration of biotin-BSA-Gd-DTPA.10

During MRI measurements, anesthetized mice were placed supine with the tumor located above the center of the MRI surface coil embedded in a Perspex board. The skin was stretched in order to form a smooth layer on the Perspex and the mice were immobilized using adhesive tape. Mice were covered with a paper blanket in order to reduce temperature drop during the experiment. MRI experiments were performed on a horizontal 4.7 T Bruker (Karlsruhe, Germany) Biospec spectrometer using an actively radiofrequency-decoupled 1.5 cm surface coil and a birdcage transmission coil. Vascular maturation was studied using BOLD contrast, namely, from change in MR signal intensity in response to hypercapnia (comparing inhalation of air to 95% air + 5% CO2) as described previously.4, 36 Briefly, different gas mixtures were applied to the mice via a home-built mask. Nine sequential T*2-weighted gradient echo images were obtained [flip angle 40°, no radiofrequency spoiling, slice thickness 1 mm, TR = 230 msec, TE = 10 msec, spectral width 50,000 Hz, FOV 3.5 cm, 256 × 256 pixels, in-plane resolution 136 μm, 117 sec/image, no dummy scans; 2 averages were acquired for each image; signal-to-noise ratio (SNR) of 65]. The first image following gas transition was discarded in data analysis to allow time for complete gas exchange. Apparent permeability surface-area product (APS) was studied using DCE-MRI. The contrast agent, biotin-BSA-Gd-DTPA, was synthesized as previously reported.9 This contrast material is large enough to remain intravascular in most nonangiogenic vessels, it has long lifetime in circulation and it allows detection in histologic sections for validation of the MRI-derived APS38 and high-resolution analysis of the distribution of the extravasated material.37 An intravenous dose of 12 mg/mouse in 0.2 ml was injected via tail-vein catheter. T1-weighted spin echo images were obtained (TR 100, 200, 500, 1,000 msec; TE 10.6 msec; spectral width 50,000 Hz; FOV 3.5 cm; slice thickness 1 mm; matrix 128 × 128 zero-filled to 256 × 256; acquisition time 52 sec for each time point; 2 averages were acquired for each image; total of 4 precontrast and 16 postcontrast images; TR 200 msec; TE 10.6 msec consecutive images were measured after administration of the contrast material). At the end of the MRI scanning, the mice were euthanized by anesthesia overdose and the tumor was retrieved for histologic analysis.

Analysis of MRI data

MRI data were analyzed on O2 workstation (Silicon Graphics, Mountain View, CA) using MATLAB (MathWorks, Inc., Natick, MA). Vascular maturation maps were generated using pixel-by-pixel t-test.4 Signal intensity acquired during inhalation of air and air-CO2 was compared using t-test (2-tails unpaired). Pixels with nonsignificant changes (p > 0.05) and data sets showing motion were discarded. Maps of the apparent permeability surface area product (APS) were generated as described previously.9 Briefly, precontrast R1 relaxation maps were generated using the variable TR images. The R1 maps were used for calculation of concentration maps of biotin-BSA-Gd-DTPA from each of the postcontrast spin echo images. These concentration maps were used for derivation of APS maps, the initial rate of contrast accumulation (slope was derived from linear fit of the first 16 min; slopes with r < 0.4 were discarded). Maturation and APS maps were color-coded (blue and green, respectively) using Adobe Photoshop 6 (Adobe Systems, Inc., San Jose, CA).


Tumors were fixed in Carnoy solution (ethanol:chloroform: acetic acid 6:3:1). Four micrometer paraffin-embedded sections were deparaffinized with xylene for 5 min followed by sequential hydration with 100%, 95% and 70% ethanol and double-distilled water for 5 min each time. Sections were washed with PBS for 5 min followed by overnight blocking of nonspecific binding using 1% BSA in PBS at 4°C. The sections were then incubated simultaneously with diluted 1:30 monoclonal antibodies against α-smooth muscle actin (α-SMA) conjugated to alkaline phosphatase (Sigma Chemical, St. Louis, MO) in PBS + 1% BSA and avidin-FITC conjugate (Sigma; for the biotinylated-BSA) at room temperature (RT) for detection of the MR contrast material. Specimens were washed twice in PBS and incubated with FAST RED (Sigma) until color developed. The slides were washed with distilled water and tap water, counterstained with Mayer's hematoxylin solution (Sigma) and sealed with a coverslip using Vectashield (antifade mounting with DAPI; Vector Laboratories, Burlingame CA). The slides were examined by Optiphot2 microscope (Nikon, Tokyo, Japan) and photographed by CCD camera (DVC, Austin, TX).


Total RNA was isolated using Tri-Reagent (Molecular Research Center, Cincinnati, OH) according to the manufacturer's instruction and reverse-transcribed in 20 μl volumes using RNase H reverse transcriptase (Super-Script II; Invitrogen Life Technologies, Carlsbad, CA) with 180 picomol hexamer random primer. Aliquots (2 μl) of the reverse transcription products were used for PCR. The following sense and antisense primers were used (Genbank accession numbers in parentheses): hu-VEGF (M32977), nucleotides 358–379 and 962–980, hu-Ang-1 (U83508) nucleotides 400–424 and 1164–1186, mu-Ang-1 (U83509) nucleotides 398–421 and 1161–1183, hu-Ang-2 (AF004327) nucleotides 775–796 and 1598–1621, mu-Ang-2 (AF004326) nucleotides 637–658 and 1459–1482. PCR parameters were 2 min at 94°C, 5 cycles of 30 sec at 94°C, 60 sec at 55°C, 90 sec at 72°C, followed by 25 cycles of 30 sec at 94°C, 60 sec at 60°C, 90 sec at 72°C and 10 min at 72°C.

In situ hybridization

Probes for in situ hybridization were prepared by RT-PCR using the following sense and antisense primers: mu-Ang-1 accession number MMU83509, nucleotides 1162–1182 and 1262–1280; mu-VEGF accession number NM009505, nucleotides 61–79 and 238–255 nucleotide; sequence was verified and fragments (119 bp for Ang-1 and 194 bp for VEGF) were cloned into the pGEM-T Easy vector system (Promega, Madison, WI). These probes were designed to share high degree of homology with human cDNAs (hu-Ang-1 accession number NM001146 and hu-VEGF accession number AF486837). Digoxigenin-labeled riboprobes were produced by in vitro transcription using a digoxigenin RNA labeling kit (Roche, Mannheim, Germany). Paraffin sections of MLS tumors were deparaffinized in xylene (3 × 5 min), taken through a graded series of ethanols (1 × 2 min in 100, 95% and 70%) and washed in TBS (Tris-buffered saline, pH 7.5). Proteinase K (Sigma) digestion was carried out at 42°C for 20 min at a concentration of 20 μg/ml in a TBS buffer containing 1 mM EDTA followed by postfixation in 4% paraformaldehyde in PBS. After 2 TBS rinses, the sections were dehydrated through graded ethanols (70%, 90% and 100%) and air-dried. Slides were then preincubated with hybridization mixture [2 × standard saline citrate (SSC), 10% dextran sulfate, 1 × Denhardt's solution (Sigma), 50% formamide and 0.02% SDS] in a humidified oven for 30 min at 42°C. Hybridization was initialized by addition of digoxigenin-labeled antisense or sense riboprobes (1 ng/ml), yeast tRNA (100 μg/ml, Sigma) and carried out overnight at the above conditions. At the end of incubation, slides were rinsed (2 × 10 min, RT) in 2 × SSC containing EDTA (1 mM), once (1 hr, 50°C) in 0.2 × SSC containing EDTA, twice (10 min, RT) in 0.5 × SSC, TBS for (5 min, RT) and TBS containing BSA (1%; Sigma) for 1 hr at room temperature. The slides were then incubated overnight in a humidified chamber with antidigoxigenin alkaline phosphatase (5 μl/ml; Roche) diluted in the above buffer and washed (2 × 5 min) with TBS at room temperature. Slides were developed using BCIP/NTB substrate kit for histochemistry (Zemed Laboratories, San Francisco, CA) according to the manufacturer's instructions.


Lack of correlation between maturation and permeability in vasculature of human ovarian carcinoma xenografts

BOLD contrast and DCE-MRI measurements were performed sequentially on the same tumors (MLS xenografted in nude mice; n = 5; Fig. 1b). Hypervascularity associated with tumor angiogenesis was evident by both methods, thus showing increased contrast enhancement and vascular permeability and increased signal changes in response to hypercapnia in the tumor and particularly in the tumor rim (Fig. 2). Spatial correlation between the vascular permeability and maturation, namely, APS derived from DCE-MRI and vascular maturation derived from BOLD contrast MRI, respectively, revealed large degree of mismatch suggesting reduced coupling between the 2 MR detectable phenotypes of angiogenesis (Fig. 2ad).

Figure 2.

Heterogeneity of vascular function and structure. (a) and (b) Anatomical T2*-weighted MR gradient echo image of 2 representative tumors (of 5). (c) and (d) Correlation map of hypercapnia-sensitive mature vessels (blue) and permeability (APS; green) from the same field of view in (a) and (b), respectively. The hypothesis of this study is that the color-coded BOLD-DCE MRI correlation maps provide molecular and phenotype biomarkers of tumor angiogenesis as shown in Figure 1(a). Histology: (e) and (f) Costaining with anti-α-SMA (red) for mature vessels and avidin-FITC for the biotinylated BSA-Gd-DTPA (green). (e) Arrows indicate high plasma leakage associated with anti-α-SMA-positive tracks/vessels; arrowheads indicate low plasma content in association with anti-α-SMA-positive tracks. (f) Arrows indicate clusters of mixed population of mature and immature (anti-α-SMA-positive and-negative staining, respectively) vessels in the tumor periphery. Arrowhead indicates mature vessels devoid of contrast material, presumably nonfunctional vessels. Scale bars = 2.5 mm (ad); 100 μm (e and f). (g) Quantitative analysis of pixel distribution from the MR color maps. Three regions of interest were sampled for each data set (5 mice), tumor (center of the tumor), rim (margin of the tumor) and control (a distinct control region). Color code: green, high permeability and low vasoreactivity to hypercapnia (indicative of low maturation); blue, low permeability and high maturation; cyan, high permeability and high maturation; black, low permeability and maturation. Asterisk, p < 0.05 (t-test, unpaired, 2-tailed; n = 5; error bars = standard deviation).

The MR findings were verified by histologic staining of sections from the same tumors (Fig. 2e and f). Double fluorescence staining was performed using antibody against α-smooth muscle actin (α-SMA; red) highlighting mature hypercapnia-sensitive vessels and with avidin-FITC (green) highlighting the MR contrast material, biotin-BSA-Gd-DTPA. The mean prevalence of the different vascular phenotypes was determined for 5 tumors (Fig. 2g). Regional mismatch between these 2 vascular parameters was detected, including mature as well as immature vessels showing low permeability in some tumor areas and high permeability in other areas (Fig. 2g). Pixels with mature hypercapnia-sensitive vessels showing high permeability (cyan) contributed approximately a third of all the mature vessels within and in the rim of the tumors, while almost no such pixels could be detected in the normal tissue (skin) within the field of view. The density of vessels as detected by MRI and in histologic sections was lower in normal skin relative to the tumor as expected in proangiogenic microenvironment of the tumor.

Infiltration of host stroma cells into tumors

Histologic sections revealed that most of the contrast agent was confined to infiltrating host stroma cell tracks transversing the tumor (Figs. 2e and 3a and b; n = 7). Most of the cells in these tracks were stained positively with anti-α-SMA antibody (Fig. 3c) and thus could be myofibroblasts, pericytes, or vascular smooth muscle cells. Costaining with avidin-FITC revealed that these infiltrating stroma tracks are associated with functional vasculature. Occasionally, capillary-like structures were observed in those tracks, and extravasated contrast material also remained largely within the stroma track areas (Fig. 3d). Thus, tumor cell islets remained avascular and extravasated macromolecular contrast material could not be detected in the interstitial space between these cells.

Figure 3.

Stroma infiltration in subcutaneous MLS tumors. (a) Eosin and hematoxylin staining of MLS tumors. (b) Higher magnification of the same tumor as in (a). Note the infiltrating stroma tracks. (c) The same tumor as in (a) stained with anti-α-SMA (red) and counterstained with hematoxylin. Note that the tracks are anti-α-SMA-positive (black arrows). (d) Triple staining anti-α-SMA (red), plasma leakage (green) and nuclei (blue) of the same field as in (c). The anti-α-SMA-positive tracks (arrow) colocalized with extravasated plasma and appear to include blood vessels (arrowheads). Scale bars = for 100 μm (a); 25 μm (bd).

Expression pattern of angiogenic factors in tumor

In order to study the expression pattern of the major regulatory genes of tumor vasculature, mRNA was isolated from tumors and reverse-transcribed. VEGF, Ang-1 and Ang-2 transcripts were detected in RNA isolated from these tumors (Fig. 4; n = 12 of 12 tumors for VEGF and Ang-1 and 9 of 12 tumors for Ang-2). However, the expression pattern was different when these MLS human ovarian carcinoma cells were cultured in vitro. Both Ang-1 and Ang-2 were not detected in monolayer culture (n = 2) or in 3D multicellular spheroid culture (pool of 15 spheroids). In contrast, VEGF165 and VEGF121 were expressed in monolayer culture; expression was further induced in 3D spheroids and was elevated in tumor xenografts (Fig. 4).

Figure 4.

Expression of proangiogenic growth factors in MLS human ovarian carcinoma in vitro and in vivo. Expression of VEGF (a) Ang-1 and Ang-2 (b) mRNA. VEGF165 and VEGF121 isoforms but not Ang-1 and Ang-2 are expressed in MLS monolayer (lane 1) and in multicellular tumor spheroids (lane 2). MLS subcutaneous tumors in CD-1 nude mice (lanes 3–5) express VEGF, Ang-1 and Ang-2.

The sequence of the VEGF and angiopoietins was determined for the tumor-derived RT-PCR products. The PCR primers for these 3 genes were designed to detect the human as well as the mouse isoforms at similar probability. Two VEGF isoforms were detected, 165 and 121, both homologues to the human gene. On the other hand, the tumor-derived Ang-1 and Ang-2 showed complete homology to the mouse angiopoietins and not to the human genes. Thus, within the tumor, angiopoietins were expressed by the infiltrating host cells (mouse), whereas VEGF was expressed the tumor cells (human).

This expression pattern was consistent with the spatial distribution of expression measured by in situ hybridization (Fig. 5). Clear compartmentalization was observed in the tumors. Islets of tumor cells, with variant size, expressed high levels of VEGF (Fig. 5ac). In contrast, the tracks of infiltrating cells (Fig. 5df) did not show any expression of VEGF but rather high levels of Ang-1, which was not detected in the tumor islets. These findings are consistent with the expression of VEGF by the human tumor cells while expression of angiopoetins was predominantly by the infiltration mouse stroma cells.

Figure 5.

In situ hybridization of angiogenic factors. Hybridization with VEGF (ac) and Ang-1 (df). (a) and (d) are higher magnification (b) and (e), respectively. (c) and (f) Antisense control probes, respectively. Note that VEGF is expressed by tumor cell islets while Ang-1 is restricted to infiltrating stroma cells.

In order to bridge the gap between the in vivo MRI analysis and the molecular analysis of gene expression, adjacent sections were used for in situ hybridization and for mapping the distribution of the biotinylated MR contrast material and the location of perivascular α-SMA-positive contractile cells (Fig. 6). Heterogeneous expression pattern of the angiogenic factors was observed as described above. VEGF expression was mainly restricted to the tumor nodules (Fig. 6a and c), while the infiltrating stroma cells express Ang-1 (Fig. 6b and d). This heterogeneous expression pattern leads to different output of vascular maturation and permeability. For example, the subregion marked with a asterisk in Figure 6(a), (c) and (e) shows mixed population of stroma and tumor cells that leads to high permeability and high maturation as consequences of both high VEGF and high Ang-1 expression. This region corresponds to the cyan regions in Figures 1(a) and 2(c) and (d).

Figure 6.

Correlation of gene expression and vascular function and structure. Hybridization with VEGF (a) and Ang-1 (b) in 2 tumors. Adjacent sections stained with anti-α-SMA (red) and counterstained with hematoxylin or double-stained with anti-α-SMA (red) and plasma leakage (green). (c) and (e) show section adjacent to (a); (d) and (f) show section adjacent to (b). The region marked with asterisk is an example for heterogeneous tissue composition, including stroma and tumor cells, which shows mixed expression of VEGF and Ang-1 and mixed pattern of vascular parameters.

The spatial heterogeneity in expression of VEGF and angiopoietins suggests that endothelial cells in different locations in the tumor will experience different concentrations of these factors. This heterogeneity is consistent with the uncoupling between vascular permeability and vascular maturation within the tumor as suggested by the functional MRI analysis of permeability and vasoreactivity.

Multiparameter analysis showed colocalization of Ang-1 expression with the α-SMA-positive infiltrating stroma cells, demonstrating their contribution to the expression of this growth factor. Interstitial as well as intravascular distribution of the MR contrast material followed closely the tracks of these infiltrating stroma cells and was almost completely excluded from the tumor cell nodules. VEGF was expressed predominantly by tumor cells in regions that showed no expression of either α-SMA or Ang-1 and almost no infiltration of the MR contrast material.


The role of stroma cells in angiogenesis and growth of tumors in general and in ovarian carcinoma in particular was the focus of many studies over the last years.29, 30, 32 The contribution of stroma cells can either be direct, through association with endothelial cells, providing structural and functional support, or indirect, by providing survival signals, thus relieving endothelial cells from their dependence on secreted growth factors for survival. Perivascular supporting cells include pericytes and vascular smooth muscle cells, as well as fibroblasts and myofibroblasts that possibly serve as pericyte progenitors. Moreover, infiltration of myofibroblasts was observed to precede and lead infiltrating endothelial sprouts and thus may be necessary also for initiation and not only subsequent stabilization of new blood vessels.32

In addition, stroma cells can support tumor angiogenesis by contribution to the expression of proangiogenic growth factors. Expression of angiogenic growth factors by tumor and stroma cells was studied in a number of systems, showing significant impact to angiogenic growth factors derived from stroma cells.29, 30

We have previously reported that initiation of growth in dormant ovarian carcinoma spheroids depends on infiltration of myofibroblast stroma cells.34, 35 In the study reported here, we further demonstrate that functional vasculature within the ovarian carcinoma tumors was colocalized with the infiltrating stroma tracks. Furthermore, we have found that stroma cells exclusively expressed angiopoietin-1 and -2, while VEGF was expressed by the tumor cells. Accordingly, VEGF was expressed by the tumor cells also in vitro in monolayer culture and expression was induced in 3-dimensional spheroid culture, while angiopoietin-1 and -2 were not expressed in vitro but expression was measured in vivo. Human versus mouse matching of the sequence allowed independent validation for the tumor-derived VEGF (human) and stroma-derived angiopoietins (mouse) in this tumor xenograft model. A similar approach was applied recently for analysis of tumor versus host expression of angiogenic growth factors in ovarian carcinoma xenografts.30

The impact of angiogenic growth factors on vascular phenotype is multifaceted. VEGF expression by the tumor can be responsible for inducing both hyperpermeability as well as endothelial cell sprouting, invasion and capillary formation. However, the impact of elevated expression of VEGF on vascular remodeling depends on the expression of angiopoietin-1 or -2. Angiopoietins have a dual role in vascular remodeling. Ang-1 promotes vascular maturation, namely, the recruitment of periendothelial cells,16 while ang-2 acts as an antagonist for Ang-1 and allows vascular destabilization. In the presence of VEGF, Ang-2 will induce angiogenesis, while in the absence of VEGF, Ang-2 will induce programmed cell death in endothelial cells leading to vascular regression.18 Permeability of blood vessels is also affected by the ratio of VEGF to Ang-1. Ang-1 was suggested to block vascular permeability induced by VEGF,15 whereas Ang-2 was not reported to antagonize this function.

The uncoupling in expression of VEGF and angiopoietins and the different spatial distribution in expression imply that endothelial cells residing in different locations of the tumors will be exposed to a wide range of concentration of either one or all growth factors. We thus hypothesized that the uncoupling in expression of the various growth factors due to independent regulation of expression in tumor and stroma cells will result in heterogeneity in vascular permeability and maturation (Fig. 1). To test this, tumors were subjected to 2 types of MRI protocols. Vasoreactivity with hypercapnia, assessed by BOLD contrast MRI, was used as a functional assay for the degree of vascular maturation, while permeability was measured from extravasation of macromolecular contrast material. Fluorescent staining of the MR contrast material and staining for smooth muscle actin provided histologic validation for the same tumors. Both MRI and histology showed the large heterogeneity in vascular phenotype and the loss of coupling between maturation and reduced permeability.

Multiparameter mapping of molecular, structural and functional aspects of the vasculature is a major technologic challenge. In particular, since each methodologic approach detects a different aspect of the vascular bed, the information obtained is expected to provide complementary information, with only partial overlap, and thus can help but cannot secure validation.39 Due to the complexity in development of new blood vessels, it is clear that under pathologic conditions, there would be divergence between different aspects that affect vessel structure and function. In the study reported here, vasoreactivity to hypercapnia was used as a functional biomarker for vascular maturation, whereas histologic labeling with α-smooth muscle actin was used as a structural marker defining the recruitment of cells with contractile capacity. Vascular patency was measured by extravasation of a macromolecular contrast material that was doubly labeled with Gd-DTPA and with biotin to allow detection both by MRI and by histology.

We have previously reported the correlation between permeability detected by MRI and that detected in histologic sections38 as well as the correlation between MRI, histology and ICP-MS analysis of the distribution of the contrast material in tumors and normal organs.37 MRI provides in vivo noninvasive and nondestructive view of the vessels and thus within the limits of sensitivity and spatial resolution can allow registration of the measured functional parameters. However, the low spatial resolution of MRI can lead to partial volume averaging of regions, including heterogeneous vascular phenotypes. Histology provides significantly superior sensitivity and spatial resolution but is destructive and correlation can be done only on adjacent slices. In the data presented here, histologic data demonstrated that the heterogeneous vascular phenotypes detected by MRI relate to actual presence of multiple vascular phenotypes in the tumor and are not due to partial volume averaging due to the low resolution of MRI. Histologic analysis of the distribution of the MR contrast material and the distribution of contractile cells in the tumor were correlated with the expression of specific angiogenic growth factors as revealed by in situ hybridization, thus bridging the gap between the molecular signals and the in vivo functional MRI data.

As expected, tumors showed increased vessel density, most but not all of which showed the classical angiogenic phenotype, namely, were immature as well as hyperpermeable. Mature vessels showing high permeability are an interesting subset of vessels, since maturation of neovasculature in normal tissues is generally accompanied by suppression of vascular permeability. Indeed, in normal skin within the MRI field of view, pixels showing mature hyperpermeable vessels are scarce. In contrast, MRI-derived correlation maps of vascular maturation and permeability showed a significant fraction of such vessels within and in the rim of tumors, accounting for approximately a third of all mature vessels within these regions.

Subcutaneous tumor xenografts only partially captures the complexity of tumor angiogenesis. However, the intimate involvement of infiltrating stroma fibroblasts and contractile myofibroblasts, and their association with the tumor vasculature, as detected in the model reported here, is typical of many human tumors. A large number of antiangiogenic therapy strategies for targeting endothelial-specific receptors are being developed, and the first agent targeting VEGF was recently approved for clinical use.40 The phenotype of the vasculature is expected to determine the susceptibility to therapy, and phenotype might alter during such treatment, even if tumor size will remain unchanged. Thus, noninvasive measures of the vasculature that can report the activity of specific signaling pathways are important. The ability of MRI to generate permeability-maturation correlation maps, which can serve as biomarker for VEGF-angiopoietin signaling, can help in preclinical and clinical evaluation of such therapies. Specifically, we demonstrated here the lack of spatial overlap between the 2 imaging approaches, each of which captures a different characteristic of the blood vessels.

In summary, the study reported here demonstrates the contribution to expression of angiogenic growth factors by the tumor cells and by tumor-associated stroma cells. Specifically, VEGF was expressed by the human ovarian carcinoma cells, while both Ang-1 and Ang-2 were expressed by infiltrating mouse stroma cells. Vascular heterogeneity is expected in this case due to the spatial segregation being a source for the various growth factors. Such heterogeneity was indeed observed, showing a wide range of phenotypes of maturation and permeability as detected both by histology and by MRI. In vivo monitoring of both maturation and permeability can be achieved by BOLD contrast and DCE-MRI, each of which provides a different and independent characterization of the tumor vasculature. Dynamic analysis of vascular maturation and permeability would allow the study of the intricate regulation of vascular remodeling and could possibly provide, in some cases, in vivo biomarkers for expression of VEGF, Ang-1 and Ang-2.