Dynamic Remodeling of the Vascular Bed Precedes Tumor Growth: MLS Ovarian Carcinoma Spheroids Implanted in Nude Mice

Abstract

The goal of this study was to monitor the vascular bed during the lag phase in growth of implanted spheroids as a model of tumor dormancy. Vascular development and tumor growth were followed up by magnetic resonance imaging in a model system of MLS ovarian carcinoma spheroids implanted subcutaneously in female nude mice. Apparent vessel density in a 1-mm rim surrounding the spheroid was evaluated by gradient echo imaging as a measure of the angiogenic potential of the tumor. Vascular functionality and maturation were assessed by signal intensity changes in response to hyperoxia (elevated oxygen) and hypercapnia (elevated carbon dioxide), respectively. Tumor growth was delayed by 12 to 57 days after implantation. During this long period in which tumor volume did not change, up to 6 cycles of vascular development and regression were observed. We propose here that dynamic remodeling of the vascular bed may precede exit of tumors from dormancy. The sustained oscillations in the angiogenic response to the implanted spheroid are consistent with hypoxic regulation of vascular endothelial growth factor (VEGF), combined with the role of VEGF as an essential survival factor for newly formed blood vessels. Vascular maturation, manifested by physiological vasodilatory response to carbon dioxide, may be important for conferring vascular stability and exit from dormancy.

Introduction

Malignant transformation does not result invariably in unconstrained tumor growth. In fact, frequently tumors remain dormant for many years. Thus, the frequency at which microtumors are detected in random screening is significantly higher than the incidence of clinical disease (1). Elucidation of the causes for tumor dormancy and the conditions that result in exit from dormancy are critical for understanding cancer progression.

Three major mechanisms were suggested as possible regulators of tumor dormancy. The control of the cell cycle in tumors was suggested to be associated with tumor dormancy (2). The microenvironment of the avascular tumor was suggested to increase genetic instability (3) and to promote selection of cells deficient in tumor suppressor genes such as P53 (4). The second mechanism suggests that suppression of tumor progression is conferred by the immune system (5). Suppression of angiogenesis was also associated with the phenomenon of tumor dormancy (6,7).

Exit of tumors from angiogenesis-dependent dormancy can be initiated by genetic selection for proangiogenic activity (8). Tumor growth can be promoted by the angiogenic response induced by a proximal injury (9). In epithelial ovarian tumors we demonstrated induction of tumor vascularization in response to elevated systemic levels of gonadotropins in ovariectomized mice (10). Sham-treated mice showed prolonged periods of dormancy and reduced angiogenic activity. The hormonal effect was attributed to increased expression of vascular endothelial growth factor (VEGF) in gonadotropin-induced tumor cells (10).

It must be noted, however, that hypoxia was shown to be a dominant regulatory mechanism controlling VEGF in many normal tissues and most tumors (11,12). Hypoxia appears as a consequence of uncontrolled cell proliferation in any tumor beyond the diameter of 0.5 mm (13). Transformation and unconstrained tumor cell proliferation should lead to hypoxic secretion of proangiogenic VEGF. In face of the prevalence of hypoxic control of VEGF, it is surprising that tumor dormancy can relate to suppressed angiogenesis.

We propose here dynamic rather than static control of angiogenesis during dormancy as a possible mechanism to suppress tumor progression despite the contribution of hypoxia-induced angiogenesis. To test this hypothesis we followed angiogenesis induced by multicellular tumor spheroids implanted subcutaneously in nude mice. Hypoxic regulation of VEGF expression in spheroids was documented previously (10,12,14,15). Changes in vessel density in the normal skin surrounding the implanted spheroid served as a measure of the angiogenic potential of the spheroid.

Vascular remodeling was followed noninvasively by magnetic resonance imaging (MRI). Paramagnetic deoxyhemoglobin in blood vessels broadens the resonance of the water proton signal (16,17). In gradient echo images, this effect leads to signal changes, which reflect changes in blood oxygenation or blood volume fraction. Increased blood volume fraction determined by MRI using this approach correlated with optical image analysis of vessel density for a number of angiogenic stimuli (18). Signal changes in response to hypercapnia (95% air, 5% carbon dioxide) or hyperoxia (95% oxygen, 5% carbon dioxide) were used for physiological assessment of vascular functionality and maturation (19).

By using MRI of MLS human epithelial ovarian carcinoma spheroids implanted in nude mice, we show the extensive remodeling of the vasculature, including angiogenesis and vascular regression, that precedes the initiation of tumor growth.

Materials and Methods

Spheroid Culture

MLS human epithelial ovarian carcinoma cells (10) were cultured in α-Eagle's minimum essential medium MEM supplemented with 10% fetal calf serum and antibiotics: 50 U/mL penicillin, 50 µg/mL streptomycin, and 125 µg/mL fungizone (Biolab Ltd, Jerusalem, Israel). Aggregation of cells into spheroids was initiated by plating cells from confluent cultures in agar-coated plates. After 48 hours, the spheroids were transferred to a 250-mL spinner flask (Bellco, Vineland, NJ) at spinning rate of 80 rpm. One week later the spheroids were transferred to a 500-mL spinner flask, where every 96 hours for approximately 1 month the medium was changed, and a mixture of 95% air and 5% carbon dioxide was blown over the medium for 5 minutes. Other details of spheroid culture were as reported previously (10).

Animal Protocol

Female, CD1-nude mice (12 weeks old, 25-g body weight) were anesthetized by intraperitoneal (IP) injection of ketamine (75 mg/g) and xylazine (3 mg/g) and placed in a sterile laminar flow hood. A single MLS human ovarian carcinoma spheroid, 1.2 mm in diameter, was implanted subcutaneously in the lower back through a 4-mm incision with Teflon tubing, as reported previously (18,20). The incision was formed with fine surgical scissors and closed with cyanoacrylate (Super Glue-3, Loctite, Ireland).

MRI Measurements of Vessel Density

Vessel density was determined by MRI with the intrinsic contrast provided by paramagnetic deoxyhemoglobin in vessels, which shortens T2* relaxation and thus reduces signal intensity (16,18,20). Gradient echo images were acquired on a horizontal 4.7-T Bruker-Biospec (Karlsruhe, Germany) spectrometer with an actively radio frequency decoupled surface coil, 1.5 cm in diameter, imbedded in a Perspex board, and a bird cage transmission coil. Mice were anesthetized (75 mg/kg ketamine and 3 mg/kg xylazine IP) and placed supine, with the tumor located at the center of the surface coil. Apparent vessel density was determined from gradient echo images (slice thickness of 0.55 mm, repetition time 230 ms, echo time 10 ms, and 256x256 pixels in plane resolution of 117 µm and 8 averages per mouse). The apparent density of the blood vessels (AVD), attracted by the tumor was determined as reported previously (18,20). Briefly, AVD=-ln(S/S₀) where S is the signal intensity in a ring of 1 mm surrounding the implanted spheroid, and S₀ is the intensity of signal in a control region about 7 mm from the spheroid. Previous studies showed that AVD determined by MRI showed significant correlation (r = 0.905, P = 0.0001) with the density of vessels determined by optical image analysis (18).

MRI Analysis of Vascular Function and Maturation

Functionality and maturation of the neovasculature were determined from gradient echo images acquired during inhalation of air, air-carbon dioxide (95% air, 5% carbon dioxide) and oxygen-carbon dioxide (95% oxygen, 5% carbon dioxide; carbogen), as reported previously (19). The different gas mixtures were applied to the mice face with a home-built mask. Four images were acquired at each gas mixture (117 seconds per image, slice thickness of 0.55 mm, TR=230 ms, spectral width of 30,000 Hz, field of view of 3 cm, 256x256 pixels, in plane resolution of 117 µm, TE=10 ms, 2 averages). Other experimental details were as reported previously (18,19).

Data Analysis

MRI data were analyzed on an Indigo-2 work station (Silicon Graphics, Mountain View, CA) with Paravision software (Bruker) and Matlab (The Math Works Inc, Natick, MA). Changes in signal intensity between air and air-carbon dioxide were used for derivation of vascular dilation: (VD = ln(I_{air - CO2})/I_air)/(TE C_MRI). Signal changes between air-carbon dioxide and oxygen-carbon dioxide were used for analysis of vascular function: (VF = bΔY = ln(I_oxygen-CO2)/I_air-CO2)/(TE C_MRI) (21). Where I_air, I_air-CO2 and I_oxygen-CO2 are the mean signal intensity during inhalation of air, air-carbon dioxide and oxygen-carbon dioxide respectively; TE is the echo time; Y is the fraction of oxyhemoglobin; b is the volume fraction of blood; and C_MRI=599 s^-1 at 4.7 T (18). VF measures the capacity of erythrocyte-mediated oxygen delivery from the lungs to each pixel in the image (18). Signal intensity changes due to hypercapnia were found to be predominantly the result of a change in the apparent T1 relaxation time due to a change in blood flow, whereas the signal change due to hyperoxia was caused by a change in T2* resulting from a change in blood oxygenation (Dafni and Neeman, unpublished data).

Results

Multicellular spheroids derived from MLS human ovarian epithelial cancer (1.2 mm in diameter) were implanted subcutaneously in CD-1 female nude mice as a model of tumor dormancy. Up to 22 MRI measurements of nine mice were performed during a period of 87 days. Follow-up of each mouse was terminated when the tumor reached a volume of 30 mm³.

Kinetics of Tumor Growth

Tumor growth, monitored by MRI, showed two distinct phases. The first is a lag phase, which extended during the first 12 to 57 days after implantation. During this period no significant change in tumor volume was observed. The second phase was characterized by exponential or Gompertz growth, which indicate exit from dormancy (Figure 1A and B, representative data). We found no correlation between the rate of tumor growth in the exponential phase and the duration of the lag period (r=0.63, P=0.27, n=6).

Figure 1.

Kinetics of tumor growth and vascularity. MLS spheroids, 1.2 mm in diameter, were implanted subcutaneously in female CD-1 nude mice at the age of 12 weeks. Up to 22 measurements were taken over 87 days for each mouse. (A, B) Lag phase and exponential growth of two different tumors. (C, D) Oscillations in AVD observed during the lag phase. Data were derived from analysis of gradient echo MRI as reported previously (18).

Oscillatory Changes in Vessel Density

Vascular recruitment induced by the tumor was traced noninvasively by MRI. Oscillatory pattern was detected in the vascular rim surrounding the tumor during the lag phase (Figure 1C and D). Between one to six episodes of vascular growth and regression occurred before initiation of tumor growth (Figure 2A; n=9). The exit from dormancy was accompanied by reduction in vascular density. Figure 2B represents the duration of the lag phase of each mouse (n=9). A weak correlation was found between the duration of the lag period and the number of vascular growth and regression cycles (r=0.68; P=0.03; Figure 2B).

Figure 2.

Oscillations in vascularity during the tumor lag phase. (A) Summary of number of oscillations in the lag phase (n=9). (B) Correlation between the duration of the lag phase and the number of oscillations (r=0.68, P=0.03, n=9).

Changes in Vessel Maturation and Function

The cycles of vascular development and regression preceded the exit of the implanted spheroids from dormancy. We checked whether vascular stabilization by recruitment of smooth muscle cells (SMC) and pericytes occurs during this period and is associated with exit from dormancy. MRI signal change due to vasodilatation in response to elevated levels of carbon dioxide was used here for mapping mature blood vessels (Figure 3C). The density of functional blood vessels was mapped by the change in signal intensity in response to hyperoxia (Figure 3B). Intense vascular functionality (Figure 3B) characterized the neovascularization surrounding the tumor during the lag phase. Maturation of the vessels in the vascular rim surrounding the tumor was observed, (Figure 3C), which implies that the neovasculature undergoes maturation through recruitment of pericytes and SMC before the initiation of tumor growth.

Figure 3.

Analysis of vascular function and maturation. Images were acquired 18 days after spheroid implantation and 3 days before the initiation of tumor growth. (A) Gradient echo image. The tumor in the center of the image is surrounded by a hypointense rim corresponding to neovasculature induced by the tumor. Gradient echo images were acquired during inhalation of air, air-carbon dioxide (95% air, 5% carbon dioxide) and carbogen (95% oxygen, 5% carbon dioxide). (B) Vascular functionality was assessed by signal changes in response to hyperoxia (VF). (C) Changes in signal intensity between air and air-carbon dioxide, corresponding to vasodilation (VD).

Discussion

The goal of this study was to address the conflict between the concept of angiogenesis-controlled dormancy and the prevalence of stress induced angiogenesis, which would promote vascular development toward all tumors that grow beyond 0.5 mm and develop central hypoxia. We showed here dynamic oscillations of angiogenesis and vascular regression during tumor dormancy, defined as a period in which the tumor did not grow. Our results suggest that exit from dormancy may be related to vascular remodeling and vascular stability, rather than to the initiation of vascular sprouting.

Kinetic analysis of growth and vascularization of implanted MLS ovarian carcinoma spheroids revealed extensive dynamic remodeling of the vasculature, including vascular growth and regression, which preceded the initiation of tumor growth. The reduced vessel density at the tumor rim, which was consistently observed upon initiation of tumor growth, is probably related to tumor growth so as to engulf the vessels in the rim. Alternatively, it could reflect vascular trimming resulting in improved tumor perfusion and elevated oxygenation, as suggested previously (22,23). The correlation measured between the lag and the number of cycles of vascular growth and regression suggests that the remodeling of the vasculature is not a fixed prelude period to initiation of tumor growth, but rather is related to the lag period itself.

The vascular bed in the tumor rim included functional vessels as manifested by MRI signal change in response to hyperoxia. Vascular maturation in the tumor periphery was revealed by vasodilation in response to carbon dioxide. The high density of mature vessels surrounding the implanted spheroid implies that these are not pre-existing mature vessels, but rather tumor-induced neovasculature that underwent maturation. Recruitment of vascular smooth muscle cells could play an important role in stabilizing the blood vessels in the tumor rim and preventing their regression.

Based on these results we would like to suggest the following model of vascular remodeling induced by the implanted spheroid (Figure 4). Hypoxic inner cell layers in the implanted spheroid show elevated expression of VEGF (10,12,14). In the early stages of tumor vascularization, neovascularization results in reduced expression of VEGF due to relief from hypoxia (12) and thus exposes the newly formed capillaries to VEGF withdrawal. VEGF withdrawal results in endothelial cell detachment and anoikis, specifically in neovasculature (24). Endothelial cells undergoing apoptosis display procoagulation factors (25). Thus VEGF withdrawal will lead to reduced vascular function (19,26) and secondary tumor hypoxia. This will create cycles of vascular growth and regression (Figure 4). The exit from this cyclic behavior can be related to vascular maturation, namely, the recruitment of pericytes and SMC, which would render the vessels resistant to VEGF withdrawal (27,28). The extensive remodeling of the vasculature may also select specific vascular patterns so as to improve perfusion (22,23).

Figure 4.

Model of dynamic vascular remodeling during tumor dormancy. Tumor growth leads to hypoxia, thus inducing expression of VEGF. This however does not lead inevitably to tumor exit from dormancy. The neovasculature can restore oxygenation and suppress VEGF expression. Acute VEGF withdrawal results in vascular collapse and tumor hypoxia. Such a dynamic cycle can trap the tumor in extended dormancy. Exit of tumors from dormancy will depend on stabilization of the vascular bed either by an exogenous source of VEGF (such as a local injury) or by vascular maturation.

The model presented here points to a number of regulatory elements that should be evaluated. For example, the model predicts that modulations in tumor oxygenation and VEGF production should match the changes in vessel density for the immature blood vessels. Recent developments in noninvasive methods for mapping tissue oxygenation (29) and gene expression (30,31) suggest that the experimental tools for such studies will be available in the near future.

In summary, we propose that tumor dormancy should not be regarded as a period in which angiogenesis is dormant. Understanding the dynamics of the vasculature during dormancy may help identify the key signals that lead to tumor growth and may be important in the development of methods of intervention as a tumor prevention approach. Our results point to vascular maturation as one possible regulatory element in tumor exit from dormancy.

Abbreviations

AVD

apparent vessel density

MRI

magnetic resonance imaging

SMC

smooth muscle cells

VD

vascular dilation

VEGF

vascular endothelial growth factor

VF

vascular function