Overexpression of Thrombospondin-1 Reduces Growth and Vascular Index but not Perfusion in Glioblastoma¹ Free

Angiogenesis is considered essential for tumors to grow beyond a few millimeters in size (1). A balance between angiogenic and antiangiogenic factors in the microenvironment regulates the angiogenic processes (2). This has elicited significant interest in the use of inhibitors of angiogenesis for the treatment of solid tumors, and it has been shown that treatment with exogenous inhibitors or overexpression of endogenous angiogenesis inhibitors retards tumor growth (3). The naturally occurring angiogenesis inhibitors include TSP-1³ (4), TSP-2 (5), angiostatin (6), and endostatin (7). TSP-1 is a M_r 420,000 homotrimeric multifunctional extracellular matrix glycoprotein that plays a central role in vascular homeostasis. TSP-1 inhibits proliferation and migration of endothelial cells in vitro and inhibits angiogenesis in vivo, contributing to the normal quiescence of the vasculature (8). Overexpression of TSP-1 inhibits tumor growth and decreases the microvessel density in a variety of tumors (9, 10, 11), and TSP-1 mRNA expression is associated with better prognosis in invasive cervical cancer (12). However, these studies have not yet addressed several important issues. Is the reduced vascular index a reflection of smaller tumor size or an intrinsic property of TSP-expressing tumors? Does a reduction in tumor vascular density translate into reduced perfusion in size-matched tumors?

LDF can provide noninvasive estimates of local blood perfusion in superficial tissue (13). The principle of this technique is based on the change in wavelength (Doppler shift) of the laser light that is reflected from moving subjects, the RBCs, whereas the wavelength of light reflected from stationary subjects remains unchanged. The Doppler-shifted light is converted into an arbitrary perfusion signal, which is approximately proportional to the mean blood cell velocity multiplied by the concentration of moving blood cells within the sampling volume. Comparison of LDF with the ¹³³xenon clearance method has shown high correlation (14, 15, 16), but because the proportionality factor remains unknown for the individual experiment, the LDF measurement provides only a relative measure of perfusion.

NIRS can provide noninvasive estimates of the hemoglobin concentration because hemoglobin is a strong absorber of near infrared light (17). At 800 nm (the isosbestic point of hemoglobin), the light absorption of a tissue is proportional to the total hemoglobin concentration, disturbed only marginally by other chromophores of the tissue (18, 19). The total hemoglobin concentration reflects the blood volume because the source of hemoglobin is the RBCs, but the NIRS signal per se cannot distinguish between many vessels and fewer larger vessels. We have previously shown that NIRS can provide estimates of the degree of vascularization in solid tumors similar to the information obtained by histological vessel count analysis (20).

Using these two methods, we examined the above questions in size-matched tumors. We found that overexpression of TSP-1 significantly slowed tumor growth but did not prevent the tumor from increasing in size over time. Whereas a clear reduction in tumor vascularity was observed, we found the same perfusion in size-matched TSP-1-transfected tumors and controls, indicating that TSP-1 overexpression renders a growth-limiting vascular phenotype.

Male 8-week-old athymic nude mice (NMRI-nu/nu) obtained from M&B (Ry, Denmark) were used. The mice were kept in groups of five in laminar air-flow benches and fed sterile food pellets and water ad libitum. Body weight was regularly recorded. Institutional guidelines for animal welfare and experimental conduct were followed. When appropriate, mice were anesthetized with ketamine/xylazine (10/100 mg/kg body weight, s.c.).

Tenan et al. (11) previously described the generation of clones derived from LN-229 human glioblastoma cells, which overexpress TSP-1. In this study, we used two clones overexpressing TSP-1 (C9 and E7) and a pool of three vector control clones (A7−A9). We established the three clones (A7−A9, C9, and E7) as s.c. xenografts in nude mice and serially transplanted the tumors. Small tumor blocks of about 10 mm³ were implanted on the hind leg of anesthetized mice. Established tumors of passage 3 were used in this study.

The s.c. tumor size was measured in two perpendicular dimensions (d₁ and d₂) using a slide gauge, and the tumor volume curves V(t) were obtained according to the following formula:

Time until initiation of exponential growth of the individual tumors was determined from individual volume curves plotted semilogarithmically, as described previously (21). The time until initiation of exponential tumor growth was determined as the intersection between the best-fitting line depicting the early lag phase and the linear regression line during the exponential growth phase. The validity of this simplification was documented by high correlation coefficients (R² = 0.980 ± 0.012; mean ± SD). Tumor doubling time and time to reach 300 mm³ were derived from the linear regression line during the exponential growth phase.

The mice were anesthetized 15 min before LDF and NIRS (see below) recordings to eliminate artifacts from body movements. The experiments were performed in a temperature-controlled room (25°C ± 1°C). A micromanipulator was used to maintain a reproducible localization of the probes perpendicular to and in close contact to the skin. LDF and NIRS recordings were performed two to three times during growth and on the final day, i.e., at tumor volume = ∼300 mm³.

Laser light with the wavelength 780 nm was transmitted to the skin above the tumor by a 41°C heated custom-built probe with four integrated laser/receiver units (6-mm outer diameter; 250-μm fiber separation; time constant, 0.2 s; PF415–175; Perimed, Stockholm, Sweden). The LDF probe was calibrated in motility standard solution (250 ± 15 PU; Perimed) before each experiment. The LDF signal was recorded continuously for 3 min, and the PU was determined as the mean PU value of the stabilized plateau. The probe was moved 0.25 mm backward if the total backscatter was too low.

The NIRS instrument was custom-built (NMR-Center, University of Copenhagen, Copenhagen, Denmark), including a xenon flash as the light source (l4633 Hammamatsu, Near Infrared Spectrometer) and a photo diode (Siemens BPW21 photo diode) as the light detector. An interference filter in front of the xenon flash unit results in emission of light with a wavelength of 800 ± 10 nm. A branched (Y-shaped) light guide was placed on the skin above the tumor of the anesthetized animal and transmitted the emitted light to the tissue and the reflected light from the tissue. The light reflected from the tissue was recorded by a photodiode via a second interference filter (800 ± 10 nm). Transmitting fibers were randomly mixed, and the diameter of that part of the probe is 3.0 mm.

The NIRS instrument was calibrated before each experimental session. The calibration was stable over time. Full absorption, i.e., zero signal, was set to 100 arbitrary units of absorption, and motility standard (Perimed) was calibrated to 50 arbitrary units of absorption. The NIRS value was calculated as the median of five recordings.

Tissue from tumors in each group was fixed in 4% formalin and embedded in paraffin blocks. Sections of 4–5 μm were cut and stained with H&E. Sections were examined by light microscopy.

Tissue from all tumors was frozen in cooled isopentane. CD31 immunostaining was performed on sections from the tumors. Sections were air-dried, fixed in acetone, washed in PBS and Tris (pH 7.6)-buffered saline, and incubated with 10% normal rabbit serum for 30 min. They were then incubated with a mixture of two monoclonal rat antimouse CD31 antibodies at a dilution of 15 μg/ml overnight at 4°C. The antibodies used were clone 390 (Serotec Ltd.) and MEC 13.3 (PharMingen). Rat IgG2a (Serotec Ltd.) was used as negative control. Sections were incubated with biotin-conjugated rabbit antirat immunoglobulin (DAKO) at a dilution of 1:200 (1.5 μg/ml) for 30 min. As substrate for the alkaline phosphatase reaction, we used freshly prepared Fast red Substrate System (DAKO), followed by a 10-min wash in tap water. After this, sections were counterstained with hematoxylin and mounted with aqueous mounting media.

A cross-section of each tumor was examined. Microvessel density was counted in a ×100 power field with a 5 × 5 grid. Only CD31-stained vessels crossing the intersection points of the grid were counted. For each tumor, five random fields were counted. The microvessel density was calculated as the mean of the five counts.

The necrotic fraction was determined in the cross-section of each tumor. The tumors were divided into four groups by visual examination of the cross-sections: (a) 0–25% necrosis; (b) 25–50% necrosis; (c) 50–75% necrosis; and (d) 75–100% necrosis.

Tissue homogenate from two randomly selected tumors (size, 300 mm³) in each of the three groups was sonicated three times for a few seconds by ultrasound using a Vibra Cell 50 (Sonics & Materials). Protein concentrations were determined using the BCA protein assay reagent (Pierce). SDS-PAGE was performed with 75 μg of total protein on precast 10% NuPAGE 10% Bis-Tris gels (NOVEX). Samples were denatured at 70°C for 10 min before loading. After electrophoresis, the proteins were blotted to polyvinylidene difluoride membranes (NOVEX) using a semidry blotting device (Transblot SD BIORAD). Membranes were blocked with TBS-T containing 5% powder milk followed by incubation with mouse monoclonal antibodies raised against human TSP (Ab-4, clone A6.1; NeoMarkers). After three washes in TBS-T, the membranes were incubated with a dilution of a horseradish peroxidase-conjugated secondary antibody (code number P0447; Dako) and washed three times in TBS-T. Finally, the protein expression was detected with ECL+ (Amersham Pharmacia Biotech). Mouse monoclonal antibody recognizing α-tubulin (Sigma Chemical Co.-Aldrich) was used to correct for uneven loading on SDS-PAGE.

The distribution of time until initiation of exponential tumor growth, time to reach 300 mm³, microvessel density, tumor necrosis fraction, and LDF recordings were not consistently Gaussian. Therefore, data were shown as median value and statistical evaluations of differences between treated versus untreated groups of animals using a two-tailed Mann Whitney U test for nonparametric data.

Statistical evaluations of differences in tumor doubling time, final tumor volume, and NIRS recordings between treated and untreated groups of animals were performed using an unpaired two-tailed t test. These data were shown as the mean value ± SD.

The production of TSP-1 by the serially transplanted tumors (passage 3) was confirmed by Western blot analysis of the tumor tissue. The transfected clones overexpressed TSP-1 compared with controls, [C9 overexpressed TSP-1 by 6-fold, and E7 overexpressed TSP-1 by 15-fold (Fig. 1)].

The growth of serially passaged LN-229 glioma xenografts overexpressing TSP-1 was significantly inhibited. Although no significant difference in time to initiation of tumor growth was observed, a highly significant increase in tumor doubling time after initiation of exponential tumor growth occurred in the tumors overexpressing TSP-1. The control tumors reached a final tumor size of 300 mm³, on average, within 48 days, whereas the transfected tumors eventually reached the same final tumor size after 96 days (C9) or 68 days (E7) (Table 1).

The tumor perfusion estimate obtained by LDF increased during tumor growth, reaching a plateau when tumors reached a size of 50–100 mm³. When compared at their final tumor sizes, no difference in perfusion between transfected tumors and controls was observed (Fig. 2).

The blood volume estimate obtained by NIRS increased during tumor growth and reached a tumor clone-specific plateau from tumor sizes >100 mm³. When compared at their final tumor sizes, tumors derived from clones C9 and E7 had a lower blood volume estimate than controls. The difference between tumors derived from the clone with the highest TSP-1 expression (E7) and control tumors was highly significant (P < 0.001; Fig. 3).

In accordance with the NIRS estimates, the histological examination showed a significantly lower microvascular density (vascular index) in E7 TSP-1-transfected tumors compared with controls. C7 TSP-1-transfected tumors also showed lower vessel counts than controls, but the difference was not statistically significant. Small necrotic foci were found in the control tumors, whereas larger areas of necrosis were detected in the TSP-1-overexpressing tumors of both clones (Table 2). In general, the vascular organization was less chaotic in the less vascularized TSP-1 overexpressors with consistently larger lumen areas (Fig. 4).

It has previously been shown that overexpression of TSP-1 in LN-229 human glioblastoma cells in immunodeficient nu/nu mice significantly inhibited their growth and vascular density compared with control tumors transfected with vector only (11). In that study, controls and TSP-1-expressing cells were injected in opposing flanks of the animals. Consequently, all animals had to be sacrificed at 9 weeks, when the control tumors reached an average size of about 2000 mm³. Here, we further examined (a) whether reduced vessel size was a cause of reduced tumor growth or simply reflected smaller tumor size and (b) whether size-matched tumors with or without TSP-1 overexpression had equal blood volume and perfusion rates.

We found that the overall TSP-1 content in the tumors at sacrifice (300 mm³) was increased 6-fold (C9) and 15-fold (E7) relative to vector controls. In a previous study, the production of TSP-1 measured in serum-free conditioned media from clones C9 and E7 was estimated to be 16.7-fold and 28.6-fold higher than that of control vector-transfected LN-229 cells (11). A decrease in TSP-1 overexpression after serial transplantation of tumors could be due to counterselection in TSP-1-expressing tumors.

Tenan et al. (11) also found that overexpression of TSP-1 decreased the vessel density by 52%, as determined by examination of microvessel density. However, the tumors arising from TSP-1 clones were extremely small because all of the tumors were scored at the same time point (namely, the time of large control tumors), therefore one cannot exclude that the reduced vessel number was a consequence rather than the cause of a smaller tumor size. In the present study, we found approximately the same reduction in tumor vascularity in size-matched glioma LN-229 tumors and thereby confirmed that the reduced vessel number was likely the cause of a smaller tumor size. Furthermore, here we show that angiogenic suppression by overexpression of TSP1 reduces the tumor vascularity in size-matched tumors without affecting tumor perfusion.

These results concur with data from recent studies in our laboratory showing that continuous administration of the exogenous antiangiogenic compound TNP-470 also reduced tumor vascular index in size-matched tumors (20). These results establish that continuous antiangiogenic therapy, endogenous as well as exogenous, reduces tumor vascularity in tumors. Whereas this effect represents a measure of the antiangiogenic effect of the compounds tested, tumors regrew despite therapy. This likely reflects an adaptation to or even resistance to antiangiogenesis. The lower tumor doubling time in the TSP-1-overexpressing tumors indicates that some phenotypic changes have been induced. The tumors might adapt to the antiangiogenic pressure, e.g., by clonal selection or by an altered metabolic phenotype. Our results indicate that the phenotypic change reflects a more transport-efficient vascular phenotype rather than a decrease in nutritive demands or tolerance toward hypoxia because we found the same perfusion in TSP-1-transfected tumors and controls despite the reduced vascularity. Normally, tumors have a chaotic vascular system with a heterogeneous blood flow (22). Therefore, selection of functional vessels can reduce the tumor blood pool without affecting the global tumor perfusion, as discussed by Gillies et al. (23) and more recently Jain (24). The distinctly different appearance of CD31-positive vascular structures in controls and TSP-1 overexpressors seems to support this interpretation.

Noninvasive reflectance NIRS and especially LDF have limited measuring depth (25). Therefore, we used small (∼300 mm³) and superficial tumors only. Both LDF and NIRS recordings increased and reached a plateau during tumor growth, an observation that most likely reflects their limited measuring volume (LDF, 50–100 mm³; NIRS, 100–200 mm³). These volumes cover a substantial part of the tumor under investigation, but we cannot rule out that the perfusion and/or blood volume in the central regions of the tumors could be different from what we found. On the other hand, parts of the signal, particularly when tumors are small, may originate from surrounding tissues. The decrease in blood volume estimate immediately after implantation and before tumor growth is due to the implantation of the avascular implant in the measuring volume, whereas the immediate increase in LDF recordings might be due to a minor inflammatory reaction to the implant. Due to intrinsic biological variation in LDF and NIRS measurements, comparisons must be made between large groups of animals, using multifiber probes with optimized configurations, under identical experimental conditions.

Others have found extensive areas of tumor cell necrosis in TSP-1-overexpressing tumors compared with controls (9, 10). This effect is thought to be due to increased thrombosis and disruption of tumor vasculature and will largely depend on whether TSP-1 is present at tumor initiation or whether an exogenous TSP treatment is applied after tumor establishment. We also observed a trend toward more necrosis in the TSP-1-transfected tumors.

In conclusion, overexpression of TSP-1 significantly inhibits tumor growth and reduces vascularity in size-matched glioma LN-229 xenografts. The continuous antiangiogenic suppression by overexpression of TSP-1 probably induces selection pressure for a more efficient vascular phenotype because TSP-1-transfected tumors have the same tumor perfusion as controls, despite the reduced vascularity. Establishing the molecular basis for this adaptation is important to find new therapeutic strategies against tumor recurrence after antiangiogenic therapy.