Recent studies have indicated that the loss of p16 is a frequent event in the progression of malignant gliomas. The loss of p16 promotes the acquisition of malignant characteristics in gliomas, which are among the most angiogenic of all human tumors. High-grade gliomas are distinguished from low-grade gliomas by intense angiogenesis in addition to their frequent loss of p16. New therapeutic strategies aimed at inhibiting tumor angiogenesis on the basis of molecular mechanisms are theoretically attractive. Here we evaluate the effect of p16 gene replacement on the angiogenesis of gliomas. Infection with a recombinant replication-defective adenovirus vector containing the cDNA of wild-type p16 significantly reduced the expression of vascular endothelial growth factor, which is thought to be a pivotal mediator of tumor angiogenesis, in p16-deleted glioma cells. Restoring wild-type p16 expression into p16-deleted glioma cells markedly inhibited angiogenesis induced by tumor cells in vivo. Furthermore, wild-type p16 inhibited neovascularization more potently than did wild-type p53 transfer. These findings indicate that the p16 gene plays an important role in the regulation of glioma angiogenesis, suggesting a novel function of the p16 gene.

Angiogenesis is an essential prerequisite for aggressive tumor proliferation and spreading (1). Prominent neovascularization and vascular endothelial proliferation are features frequently encountered in high-grade gliomas (2); thus, the interaction between glioma cells and the tumor vasculature has received considerable attention. The onset of tumor angiogenesis seems to occur at various steps induced by tumor-producing angiogenic-related molecules during the progression of malignant transformation (2, 3, 4). Among these angiogenic factors, VEGF3 plays a pivotal role in tumor angiogenesis. VEGF is a dimeric glycoprotein secreted by normal and transformed cells, and it acts as both a selective endothelial cell mitogen and a potent inducer of microvessel permeability (5, 6). Correspondingly, two VEGF receptors (Flt-1 and KDR/Flk-1) are preferentially expressed in invading and proliferating endothelial cells (4). The expression of VEGF and its receptors, which markedly contribute to tumor-associated neovascularization, seems to be correlated with the malignant transformation of gliomas and the patient prognosis (2).

The link between tumor suppressor genes and angiogenesis remains obscure. However, alteration of such genes may contribute to transformation into an angiogenic phenotype following changes in the production of cytokines, which act as angiogenic or antiangiogenic factors (7, 8). Recent experiments have shown that induction of wild-type p53 down-regulates VEGF expression and inhibits angiogenesis in several tumor cells (9, 10). Clinical investigation has also revealed that mutant-type p53 expression correlates well with VEGF expression and tumor vascularity in several cancers (11, 12).

A multistep cascade proceeding from a nonneoplastic cell to a low grade and finally to a high grade of malignancy has been molecularly characterized in gliomas (13). Mutations of p53 are frequently observed and represent consistent molecular genetic alterations in low-grade gliomas. This suggests that loss of p53 function is involved in the early phases of glioma development. In contrast, much evidence has been obtained with regard to the genetic loci and genes affected during the malignant progression of gliomas. In particular, a high frequency (greater than 50%) of homozygous p16 gene deletion has been demonstrated in gliomas and is significantly more common in high-grade gliomas (14, 15, 16). Angiogenic capacity correlates with the degree of malignancy, and loss of p16 activity frequently occurs in high-grade gliomas. However, there are no reports demonstrating that inactivation of the p16 gene promotes angiogenesis in gliomas. To explore the possibility that the p16 gene regulates glioma angiogenesis, we examined the regulatory role of the tumor suppressor gene p16 in the expression of VEGF in vitro and in the process of neovascularization in vivo. Wild-type p16 gene was introduced using the recombinant adenoviral vector Ad5CMVK-p16 in human glioma cells.

Recombinant Adenovirus and Cell Lines.

Ad5CMVK-p16, a recombinant adenoviral vector, was constructed by following a previously published procedure (17). Ad5CMVK-p16 contains the cytomegalovirus promoter, wild-type p16 cDNA, obtained from Dr. D. Beach (Institute of Child Health, London, United Kingdom; Ref. 18) as well as a SV40 polyadenylation signal in a minigene cassette, which was inserted into the E1-deleted region of the modified adenovirus 5. The adenoviruses Ad5CMV-p53 and Ad5CMV-p21, which contain the p53 and p21 gene, respectively, were kindly provided by Dr. Jack A. Roth (The University of Texas M. D. Anderson Cancer Center, Houston, TX).

The human glioma cell lines U-87 MG and SW 1783 were obtained from the American Type Culture Collection (Manassas, VA), and U-251 MG was generously provided by Dr. N. Arita (The University of Osaka, Osaka, Japan). U-251 MG and U-87 MG cells have an endogenously deleted p16 gene (19), whereas SW 1783 cells contain the wild-type p16 gene (20). The p53 gene is mutated in U-251 MG (codon 273, CGT→CAT; Ref. 21) and SW 1783 cells [codon 273, CGT→CAT; confirmed by direct sequencing using a ABI PRISM 377 DNA sequencer (Perkin-Elmer, Foster City, CA)]. However, p53 is intact in U-87 MG cells (21). U-87 MG and SW 1783 cells were cultured in Eagle’s minimum essential medium supplemented with 10% heat-inactivated fetal bovine serum, 1 mm sodium pyruvate, and nonessential amino acids. U-251 MG cells were maintained in DMEM containing 10% heat-inactivated fetal bovine serum. Cells were cultured at 37°C in a humidified atmosphere containing 5% CO2. Cell infection was carried out according to a previously described technique (22). Culture medium was used for mock infection.

Western Blot Analysis.

Cells were infected with medium alone, Ad5CMV-LacZ viral control, Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16. Twenty-four h after the infection, total cell lysates were prepared by lysing cell monolayers in plates with SDS-PAGE sample buffer [125 mm Tris-HCl, 2% sodium dodecyl sulfate, 5% mercaptoethanol, and 10% glycerine (pH 6.8)] for 1 h at 4°C after rinsing the cells with PBS. Cell lysate protein (10 μg) from each sample was measured by the BCA protein assay (Pierce, Rockford, IL) and subjected to 13.5% SDS-Tris glycerine gel electrophoresis; it was then transferred to a polyvinylidene difluoride membrane (Bio-Rad Laboratories, Hercules, CA). Membranes were blocked with 1% dry milk and 0.1% Tween 20 in Tris-buffered saline and incubated with the primary antibodies, mouse antihuman p21/WAF-1 monoclonal antibody (Oncogene Research Products, Cambridge, MA), mouse antihuman p53 monoclonal antibody (DAKO, Carpinteria, CA), and mouse antihuman p16 monoclonal antibody (PharMingen, San Diego, CA); membranes were also incubated with a secondary antibody, horseradish peroxidase-conjugated sheep antimouse IgG (Amersham, Arlington Heights, IL). The membranes were developed according to the Amersham enhanced chemiluminescence protocol.

Flow Cytometric Assay.

To analyze the DNA histogram, cell suspensions were prepared by trypsinization at 36 and 72 h after infection with Ad5CMV-LacZ viral control, Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16. Cells were fixed with 70% ethanol and incubated with RNase A (250 μg/ml; Sigma, St. Louis, MO) for 1 h at 37°C and propidium iodide (50 μg/ml; Sigma) for 10 min. Each sample was filtered through a 50-μm nylon mesh to obtain a single-cell nuclear suspension. The samples were analyzed on a FACSCalibur (Becton Dickinson, Mountain View, CA) with a minimum of 20,000 events for each sample. An argon laser at 488 nm excitation was used, and the samples were collected at 610 nm. ModFit LT (Becton Dickinson) was used for data analysis. The mean peak fluorescence was determined for each histogram.

RT-PCR Analysis.

Cells (1.25 × 105) were infected with medium alone, Ad5CMV-LacZ viral control, Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16. Total RNA was isolated from cultured cells using Trizol reagent (Life Technologies, Inc., Gaithersburg, MD) 24 h after infection. Total RNA sediment (400 ng) was reverse-transcribed in 50 mm Tris-HCl (pH 8.3), 75 mm KCl, 10 mm DTT, 3 mm MgCl2, 0.5 mm deoxynucleotide triphosphates, and 0.2 μg of oligo(dT)12–18 primer (Life Technologies, Inc.) with 200 units of Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.) at 37°C for 60 min. The PCR was performed with human VEGF- and GAPDH (as an internal standard)-specific primers. The sequence of the VEGF primer 1 was 5′-GGATGTCTATCAGCGCAGCTAC-3′, and the sequence of VEGF primer 2 was 5′-TCACCGCCTCGGCTTGTCACATC-3′ (23). This primer set can detect mRNA encoding all four molecular species of VEGF giving rise to 322-, 454-, 526-, and 577-bp bands in the case of VEGF121, VEGF165, VEGF189, and VEGF206, respectively. The sequence of GAPDH primer 1 was 5′-CAAAGTTGTCATGGATGACC-3′, and the sequence of GAPDH primer 2 was 5′-CCATGGAGAAGGCTGGGG-3′, which was expressed as a 195-bp band (24). Reverse transcription products (3 μl) were amplified by Taq DNA polymerase (Promega, Madison, WI) in 25 μl of a reaction mixture containing 5 mm Tris-HCl (pH 8.0), 10 mm NaCl, 10 μm EDTA, 0.1 mm DTT, 1.5 mm MgCl2, 0.15 mm deoxynucleotide triphosphates, and 0.2 μm VEGF or GAPDH primers. The PCR consisted of 30 cycles of 1 min at 94°C, 2 min at 61°C, and 2 min at 72°C. Products were then electrophoresed on a 1% agarose gel. The intensity of the ethidium bromide luminescence was measured using a Fluor Imager SI (Amersham). The amplified 322- and 454-bp fragments of VEGF cDNA were sequenced after subcloning into pGEM-T Easy Vector (Promega).

Quantification of mRNA by RT-PCR was done as follows. The plasmid-containing VEGF cDNA was amplified with increasing doses under the same conditions as the PCR. The products increased exponentially in a dose-dependent manner between 10−7 and 10−3 μg. The PCR product of each reaction mixture was plotted on a semilogarithmic graph, and the regression equations were determined by the linear portion of the curves. The initial amounts of the template were estimated by the standard curves. The amount of VEGF was normalized with that of GAPDH and expressed relative to that of the mock infection.

Northern Blot Analysis.

Total RNA was extracted from the culture cells using Trizol reagent (Life Technologies, Inc.) 24 h after infection, whereas mRNA was isolated from the total RNA using Oligotex-dT30 (Takara, Tokyo, Japan). mRNA (3 μg) was fractionated on a 1% denaturing formaldehyde/agarose gel and transferred to Hybond-N membranes (Amersham) by capillary elution. Nylon filters were prehybridized at 42°C for 2 h in a buffer containing 50% formamide, 4× SSC (20× SSC = 3.0 m NaCl and 0.3 m sodium citrate), 8× Denhardt’s solution, and 0.2 mg/ml sheared denatured herring sperm DNA and hybridized with the 32P-labeled VEGF probe (500,000 cpm/ml buffer) at 42°C overnight. Filters were washed twice for 5 min each at 42°C in 2× SSC and 0.5% SDS and then washed once for 30 min in 0.1× SSC and 0.5% SDS. Autoradiography was then performed using a FUJIX BAS 1000 bio-imaging analyzer (Fuji Photo Film Co., Ltd., Tokyo, Japan).

The amplified 454-bp fragment of VEGF165 cDNA was used as a VEGF probe. This probe recognizes a VEGF transcript of approximately 4.0 kb. The VEGF cDNA probe was purified by 1% agarose gel electrophoresis, recovered with GENECLEAN III (BIO 101, La Jolla, CA), and radiolabeled with [α-32P]dCTP (Amersham) using a DNA labeling kit (Amersham).

ELISA.

Cells (1.5 × 105) were plated on 12-well plates and grown for 24 h before viral infection. The cells were then infected with Ad5CMV-LacZ viral control, Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16. Culture medium was used for mock infection. The medium was changed to G-5 supplement (Life Technologies, Inc.) containing neurobasal medium (Life Technologies, Inc.) with medium alone, Ad5CMV-LacZ, Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16 after rinsing the cell monolayers with PBS 24 h after infection. After 12 h, cell supernatants were collected and stored at −80°C until use. The 96-well plates were coated with supernatants and serially diluted standard solutions (human recombinant VEGF165; R&D Systems, Minneapolis, MN) overnight at 4°C. The plates were washed six times with PBS containing 0.05% Tween 20 (washing buffer) and blocked with 5% BSA in PBS for 30 min at 37°C. After the plates were washed, goat antihuman VEGF polyclonal antibody (R&D Systems) was added and incubated for 2 h at room temperature and then washed as described above. A 1:8000 dilution of a horseradish peroxidase-conjugated rabbit antigoat IgG secondary antibody (DAKO) was added and allowed to incubate for 3 h at room temperature. The plates were then washed six times with a washing buffer. Substrate solution (0.05% o-phenylenediamine dihydrochloride and 0.3% H2O2 in 0.1 m citric acid buffer) was added to the plates. After 10 min in darkness, the reaction was stopped by the addition of sulfuric acid. Absorbance at 490 nm was determined on a plate reader (Inter Med, Tokyo, Japan), and the VEGF content of the samples was estimated from the linear portion of the standard curve. Three separate experiments were performed for each cell line.

Dorsal Air Sac Assay.

As described previously (25), each cell was suspended in PBS at a concentration of 1 × 108 cells/ml. This suspension (0.1 ml) was injected into a chamber (Millipore, Bedford, MA) consisting of a ring with a filter (pore size, 0.22 μm) on either side. This chamber contained PBS or tumor cells infected with culture medium, Ad5CMV-LacZ, Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16. The chamber was implanted into a dorsal air sac produced by the injection of 10 ml of air in the dorsum of male 6–8-week-old nude mice. Four mice in each group were sacrificed and skinned carefully on day 2. After the implanted chamber was removed from the s.c. air fascia, a ring without filters was placed on the same site and then photographed. The photographic area of the air sac fascia that showed a dense capillary network was determined using an image analyzer (Imaging Research, Ontario, Canada). The newly formed blood vessels were morphologically distinguishable from the preexisting background vessels by their zigzagging character (26).

Statistics.

All values are given as the mean ± SD. Student’s t tests were used in all analyses.

Gene Transduction Efficiency and p21, p53, and p16 Expression in Human Glioma Cells.

β-Galactosidase transduction was assessed to determine adenoviral transduction efficiency as described previously (22). All cell lines infected with Ad5CMV-LacZ over a m.o.i. of 100 exhibited 100% blue cells. Infection with Ad5CMV-p21 and Ad5CMVK-p16 at a m.o.i. of 100 and infection with Ad5CMV-p53 at a m.o.i. of 50 had no effect on cell growth until 36 h after infection and maintained a 100% viability that was determined by trypan blue exclusion in the examined cells (data not shown). With these results, Ad5CMV-p21 and Ad5CMVK-p16 at a m.o.i. of 100 and Ad5CMV-p53 at a m.o.i. of 50 were used in this study to exclude the possibility that the change in VEGF expression is attributed to growth inhibition induced by gene transduction.

Western blot analysis was performed to confirm p21, p53, and p16 expression in virally transduced cells,. Endogenous p21 was hardly detectable in U-251 MG and SW 1783 cells but was detected in U-87 MG cells. U-251 MG and SW 1783 cells exhibited high levels of mutant-type p53 protein. Endogenous p16 was absent in U-251 MG and U-87 MG cells but was detected in SW 1783 cells. In the gene transfection study, exogenous p21, p53, and p16 were clearly demonstrated 24 h after infection with Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16, respectively (Fig. 1).

Cell Cycle Analysis on Exogenous p21, p53, and p16 Protein Expression.

We performed flow cytometric analysis of the cell cycle to evaluate the effect of exogenous wild-type p21, p53, and p16 expression on the cell growth of U-251 MG, U-87 MG, and SW 1783 glioma cells (Table 1). U-251 MG cells accumulated significantly in the G0-G1 phase at 36 and 72 h after infection with Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16, as compared with infection with Ad5CMV-LacZ. G0-G1 accumulation occurred at the same levels among cells infected with Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16. In contrast, U-87 MG cells did not arrest in the G0-G1 phase of the cell cycle 36 h after infection with Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16 but significantly accumulated in the G0-G1 phase 72 h after infection with Ad5CMVK-p16, as compared with infection with Ad5CMV-LacZ, Ad5CMV-p21, and Ad5CMV-p53. SW 1783 cells did not accumulate in the G0-G1 phase at 36 and 72 h after infection with Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16.

We confirmed that no significant change in the G0-G1 phase was found among Ad5CMV-p21-, Ad5CMV-p53-, and Ad5CMVK-p16-treated U-251 MG, U-87 MG, and SW 1783 cells at 36 h after infection. Therefore, VEGF expression was assessed within 36 h after infection.

Reduction of VEGF RNA by Introduction of Wild-Type p16.

RT-PCR was performed to detect VEGF RNA expression in virally transduced cells 24 h after infection (Fig. 2). Two bands corresponding to VEGF165 and VEGF121, confirmed by sequencing, were observed in all cells analyzed. In U-251 MG cells, VEGF165 and VEGF121 RNA expression in Ad5CMV-p53-treated cells decreased as compared with that in Ad5CMV-p21-treated cells (P < 0.01 and P < 0.05, respectively). Ad5CMVK-p16 reduced VEGF165 and VEGF121 RNA expression more potently than Ad5CMV-p53 (P < 0.01). In U-87 MG cells, Ad5CMV-p53 did not affect VEGF RNA expression, but Ad5CMVK-p16 significantly reduced VEGF165 and VEGF121 RNA expression as compared with Ad5CMV-p21 (P < 0.05) and Ad5CMV-p53 (P < 0.05). In SW 1783 cells, Ad5CMV-p53 reduced VEGF165 and VEGF121 RNA expression as compared with Ad5CMV-p21 (P < 0.05) and Ad5CMVK-p16 (P < 0.05), but Ad5CMVK-p16 had no effect on the expression of VEGF RNA.

Furthermore, VEGF RNA expression in U-251 MG cells was estimated by Northern blot analysis (Fig. 3). Twenty-four h after infection with Ad5CMVK-p16, VEGF RNA expression decreased markedly as compared with Ad5CMV-LacZ, Ad5CMV-p21, and Ad5CMV-p53 infection (P < 0.01). This result was compatible with that of RT-PCR.

Suppression of VEGF Secretion Attributed to Wild-Type p16 Expression.

To measure the VEGF protein levels in a medium of cells transduced with adenoviral vectors, cell supernatants were collected 36 h after infection with Ad5CMV-LacZ, Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16. VEGF expression did not change in any of the cell lines after infection with Ad5CMV-LacZ and Ad5CMV-p21. VEGF secretion in Ad5CMV-p53-treated U-251 MG cells was significantly suppressed by 23.0% as compared with that of Ad5CMV-p21-treated cells (P < 0.05). Furthermore, Ad5CMVK-p16 reduced VEGF secretion in U-251 MG cells by 78.4% and 72.0% as compared with Ad5CMV-p21 (P < 0.0001) and Ad5CMV-p53 (P < 0.001), respectively. In U-87 MG cells, Ad5CMV-p53 had no effect on VEGF expression, but Ad5CMVK-p16 reduced the expression of VEGF by 82.5% as compared with Ad5CMV-p21 (P < 0.001). In SW 1783 cells, no significant reduction of VEGF protein levels was found between the mock-infected and virally infected cells (Fig. 4). Adenovirus-mediated overexpression of wild-type p53 significantly decreased VEGF protein expression in U-251 MG cells. In p16-deleted cell lines (U-251 MG and U-87 MG cells), overexpression of wild-type p16 suppressed VEGF production more potently than wild-type p53.

Inhibition of Tumor Cell-induced Angiogenesis in Vivo by Wild-Type p53 and p16 Gene Transfer.

We examined the development of a capillary network on the air sac fascia of nude mice (four mice/group). The photographs shown in Fig. 5 display air sac fascia in contact with a chamber containing U-87 MG cells or PBS on day 2 after implantation into the nude mice. Mock-, Ad5CMV-LacZ-, and Ad5CMV-p21-infected cells were found to induce the development of more extensive capillary networks than PBS, whereas Ad5CMVK-p16-infected tumor cells induced only a few newly formed vessels. In U-251 MG cells, the mean area of capillary networks was reduced by 34.6% in mice implanted with Ad5CMV-p53-treated cells as compared with those implanted with Ad5CMV-p21-treated cells (P < 0.05; Fig. 6). Furthermore, Ad5CMVK-p16 diminished newly formed vessels more potently than Ad5CMV-p53 (P < 0.05). Similarly, in U-87 MG cells, infection with Ad5CMV-p53 and Ad5CMVK-p16 reduced newly formed vessels by 36.1% and 66.9%, respectively, as compared with Ad5CMV-p21 (P < 0.05 and P < 0.001, respectively). Ad5CMVK-p16 decreased neovascularization more potently than Ad5CMV-p53 (P < 0.05). In SW 1783 cells, Ad5CMV-p53 inhibited angiogenesis induced by tumor cells more potently than Ad5CMV-LacZ (P < 0.05), but no significant change in neovascularization was found among treatments with Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16. Wild-type p16 transfer inhibited angiogenesis more markedly than wild-type p53 in p16-deleted cell lines.

In the present study, Ad5CMVK-p16 significantly decreased VEGF RNA and protein expression in p16-deleted cells (U-251 MG and U-87 MG cells) but did not change expression in SW 1783 cells with wild-type p16. On the other hand, Ad5CMV-p21 did not suppress VEGF expression in any of the cells. The degree of growth inhibition and cell cycle arrest did not change between p21- and p16-treated cells 36 h after infection. Therefore, we speculate that the suppression of VEGF RNA and protein expression in Ad5CMVK-p16-infected cells might not be related to cell growth or cell cycle regulatory ability. These findings indicate that the p16 gene plays an important regulatory role in VEGF synthesis in a cell cycle-independent manner. To further ascertain the effect of tumor suppressor gene transfer on angiogenesis, we used a dorsal air sac assay in nude mice. Ad5CMV-p21 had no effect on angiogenesis induced by tumor cells in vivo. Ad5CMVK-p16 significantly reduced tumor neovascularization in p16-deleted cells but not in SW 1783 cells. Furthermore, restoration of wild-type p16 markedly reduced neovascularization in p16-deleted cells as compared with the effect of wild-type p53. These results provide novel evidence that the p16 gene is implicated in more significant reduction of the tumor angiogenesis mediated by VEGF in gliomas.

The complexity of the control pathways of tumor suppressor genes involves other aspects of tumor biology (7, 8, 27). Previous reports have indicated that overexpression of an exogenous wild-type p16 in glioma cells lacking endogenous p16 expression modifies the transformed phenotype of glioma cells (19, 28). Restoring p16 activity in glioma cells significantly inhibited cell growth by arresting them in the G0-G1 phase of the cell cycle, which is closely associated with the presence of functional retinoblastoma protein (29). Moreover, combinatorial transfer of the p53 and p16 genes by adenoviral vectors led to the efficient induction of apoptosis, which could not be achieved by p53 on its own in retinoblastoma protein-positive tumor cells (30). A recent study demonstrated that p16 overexpression suppressed tumor cell invasion by an inhibition of matrix metalloproteinase 2 in glioma cells (31). Although the exact mechanism by which p16 down-regulates VEGF is unknown, it is possible that p16 down-regulation of VEGF and the resulting inhibition of tumor neovascularization would contribute to a bystander antitumor effect that may be mediated, at least in part, by an antiangiogenesis mechanism. Therefore, the anticancer effect of p16 may be modulated by different factors, such as cell cycle arrest, growth inhibition, suppression of invasiveness, and inhibition of angiogenesis.

Wild-type p53 has been noted to control tumor angiogenesis by regulating the expression of VEGF (9, 10). Our results show that the induction of Ad5CMV-p53 caused a smaller measure of angiogenesis suppression than the more prominent effect of p16 gene restoration. In U-87 MG cells, Ad5CMV-p53 did not suppress VEGF expression but significantly inhibited angiogenesis in vivo. In glioma cells, wild-type p53 probably induces the release of an unidentified inhibitor of angiogenesis. Van Meir et al.(8) demonstrated that p53 regulates the expression of glioma-derived angiogenesis inhibitory factor. Recently, a candidate glioma-derived angiogenesis inhibitory factor was identified as a p53-inducible gene product containing five thrombospondin 1 repeats specifically expressed in the brain (brain-specific angiogenesis inhibitor 1; Ref. 32). This study indicates that angiogenesis inhibitors might be up-regulated by wild-type p53 transduction.

Because conventional adjuvant therapy using surgery, radiotherapy, and chemotherapy against malignant gliomas has limitations in improving the prognosis (33), new rational and effective strategies are required to achieve a therapeutic breakthrough. Malignant gliomas are the most common primary brain tumors, which are characterized by increased mitotic activity, a highly invasive nature, and prominently neovascularized neoplasm (34). In this study, wild-type p16 transfer inhibited tumor neovascularization markedly more potently than did wild-type p53, which has been noted to inhibit tumor angiogenesis. Our results indicate that the vector-mediated transduction of p16 may have a potential role in the treatment of malignant gliomas. Recent studies (35, 36) have identified effective antiangiogenic agents such as angiostatin and endostatin that inhibit tumor growth. Combined delivery of p16 and other genes that code antiangiogenic substances may promote further tumor regression. The present study and more recent data disclose that p16 performs multiple functions as a tumor suppressor in fundamental processes during malignant transformation; such processes include cell cycle progression, tumor invasion, and angiogenesis. Although additional studies will be necessary to assess the suppression of tumor growth in vivo, wild-type p16 gene replacement may be an effective strategy in the treatment of malignant gliomas.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

      
1

Supported in part by Grants-in-Aid from the Ministry of Education, Science, Sports and Culture of Japan and the Ministry of Health and Welfare of Japan.

            
3

The abbreviations used are: VEGF, vascular endothelial growth factor; RT-PCR, reverse transcription-PCR; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; m.o.i., multiplicity of infection.

Fig. 1.

Detection of p21, p53, and p16 protein in adenovirally transduced cells by Western blot analysis. U-251 MG, U-87 MG, and SW 1783 human glioma cells were infected with medium only (mock), Ad5CMV-LacZ (lacZ), Ad5CMV-p21 (p21), Ad5CMV-p53 (p53), and Ad5CMVK-p16 (p16). Cellular extracts were prepared 24 h after infection and subjected to Western blot analysis using monoclonal antibodies against p21 (A), p53 (B), and p16 (C). Endogenous (mock-infected) and exogenous (adenovirally infected) p21, p53, and p16 proteins are shown.

Fig. 1.

Detection of p21, p53, and p16 protein in adenovirally transduced cells by Western blot analysis. U-251 MG, U-87 MG, and SW 1783 human glioma cells were infected with medium only (mock), Ad5CMV-LacZ (lacZ), Ad5CMV-p21 (p21), Ad5CMV-p53 (p53), and Ad5CMVK-p16 (p16). Cellular extracts were prepared 24 h after infection and subjected to Western blot analysis using monoclonal antibodies against p21 (A), p53 (B), and p16 (C). Endogenous (mock-infected) and exogenous (adenovirally infected) p21, p53, and p16 proteins are shown.

Close modal
Fig. 2.

Effect of Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16 on VEGF RNA expression in U-251 MG, U-87 MG, and SW 1783 cells. Cells were infected with medium alone (mock), Ad5CMV-LacZ (lacZ), Ad5CMV-p21 (p21), Ad5CMV-p53 (p53), and Ad5CMVK-p16 (p16). cDNA preparations of 454 and 322 bp corresponding to VEGF165 and VEGF121, respectively, were analyzed by RT-PCR in each cell. GAPDH transcripts were used as an internal control. A, PCR products were electrophoresed on 1% agarose gels and visualized by ethidium bromide staining. M, marker (λ-HindIII digest). B, the amount of VEGF165 (▪) and VEGF121 (□) RNA was estimated as described in “Materials and Methods.” Values are expressed as a percentage of the mock control group from four experiments. Bars, SD. a, P < 0.05 versus lacZ and p21. b, P < 0.05 versus lacZ, p21, and p53. c, P < 0.05 versus lacZ, p21, and p16.

Fig. 2.

Effect of Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16 on VEGF RNA expression in U-251 MG, U-87 MG, and SW 1783 cells. Cells were infected with medium alone (mock), Ad5CMV-LacZ (lacZ), Ad5CMV-p21 (p21), Ad5CMV-p53 (p53), and Ad5CMVK-p16 (p16). cDNA preparations of 454 and 322 bp corresponding to VEGF165 and VEGF121, respectively, were analyzed by RT-PCR in each cell. GAPDH transcripts were used as an internal control. A, PCR products were electrophoresed on 1% agarose gels and visualized by ethidium bromide staining. M, marker (λ-HindIII digest). B, the amount of VEGF165 (▪) and VEGF121 (□) RNA was estimated as described in “Materials and Methods.” Values are expressed as a percentage of the mock control group from four experiments. Bars, SD. a, P < 0.05 versus lacZ and p21. b, P < 0.05 versus lacZ, p21, and p53. c, P < 0.05 versus lacZ, p21, and p16.

Close modal
Fig. 3.

Inhibition of VEGF RNA expression by introduction of wild-type p16. Northern blot analysis was performed to detect VEGF RNA expression in U-251 MG cells 24 h after infection with culture medium alone (mock), Ad5CMV-LacZ (lacZ), Ad5CMV-p21 (p21), Ad5CMV-p53 (p53), and Ad5CMVK-p16 (p16). A, human VEGF cDNA was used as a probe. B, an ethidium bromide-stained gel scanned at the 18S and 28S rRNA bands served as a loading standard. RNA was electrophoresed on 1% agarose gels. C, VEGF RNA expression was quantitated by densitometry of autoradiograms. Signal intensities were normalized to 28S RNA levels to correct for differences in RNA loading. Results are expressed as a percentage of the mock control group from three experiments. Bars, SD. a, P < 0.05 versus lacZ, p21, and p53.

Fig. 3.

Inhibition of VEGF RNA expression by introduction of wild-type p16. Northern blot analysis was performed to detect VEGF RNA expression in U-251 MG cells 24 h after infection with culture medium alone (mock), Ad5CMV-LacZ (lacZ), Ad5CMV-p21 (p21), Ad5CMV-p53 (p53), and Ad5CMVK-p16 (p16). A, human VEGF cDNA was used as a probe. B, an ethidium bromide-stained gel scanned at the 18S and 28S rRNA bands served as a loading standard. RNA was electrophoresed on 1% agarose gels. C, VEGF RNA expression was quantitated by densitometry of autoradiograms. Signal intensities were normalized to 28S RNA levels to correct for differences in RNA loading. Results are expressed as a percentage of the mock control group from three experiments. Bars, SD. a, P < 0.05 versus lacZ, p21, and p53.

Close modal
Fig. 4.

Effect of wild-type p21, p53, and p16 expression on supernatant VEGF protein levels in U-251 MG, U-87 MG, and SW 1783 cells. Cells (1.5 × 105) in triplicate wells were incubated on 12-well plates for 24 h in 10% fetal bovine serum-containing medium with medium alone, Ad5CMV-LacZ, Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16. Each medium was changed by serum-free conditioned medium with medium alone (mock), Ad5CMV-LacZ (lacZ), Ad5CMV-p21 (p21), Ad5CMV-p53 (p53), and Ad5CMVK-p16 (p16) and incubated for another 12 h. The supernatants were assayed for VEGF using an ELISA. Values are the means ± SD of triplicate determinations. a, P < 0.05 versus mock, lacZ, and p21. b, P < 0.05 versus mock, lacZ, p21, and p53.

Fig. 4.

Effect of wild-type p21, p53, and p16 expression on supernatant VEGF protein levels in U-251 MG, U-87 MG, and SW 1783 cells. Cells (1.5 × 105) in triplicate wells were incubated on 12-well plates for 24 h in 10% fetal bovine serum-containing medium with medium alone, Ad5CMV-LacZ, Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16. Each medium was changed by serum-free conditioned medium with medium alone (mock), Ad5CMV-LacZ (lacZ), Ad5CMV-p21 (p21), Ad5CMV-p53 (p53), and Ad5CMVK-p16 (p16) and incubated for another 12 h. The supernatants were assayed for VEGF using an ELISA. Values are the means ± SD of triplicate determinations. a, P < 0.05 versus mock, lacZ, and p21. b, P < 0.05 versus mock, lacZ, p21, and p53.

Close modal
Fig. 5.

Effect of Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16 on U-87 MG cell-induced angiogenesis. In vivo angiogenesis was assessed with semipermeable membrane chambers containing glioma cells infected with culture medium alone (A), Ad5CMV-LacZ (B), Ad5CMV-p21 (C), Ad5CMV-p53 (D), and Ad5CMVK-p16 (E), and a chamber containing PBS only (F) implanted in the dorsal air sac of nude mice. The chamber allowed for the diffusion of growth factors, such as VEGF, but not of cells. Mice were sacrificed and skinned carefully on day 2. After the implanted chamber was removed from the s.c. fascia, a ring without filters was placed on the same site and photographed. The chamber ring is 10 mm in diameter. The blood vessels newly formed by glioma cells (arrowheads) were morphologically distinguishable from the preexisting background vessels by their zigzagging character.

Fig. 5.

Effect of Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16 on U-87 MG cell-induced angiogenesis. In vivo angiogenesis was assessed with semipermeable membrane chambers containing glioma cells infected with culture medium alone (A), Ad5CMV-LacZ (B), Ad5CMV-p21 (C), Ad5CMV-p53 (D), and Ad5CMVK-p16 (E), and a chamber containing PBS only (F) implanted in the dorsal air sac of nude mice. The chamber allowed for the diffusion of growth factors, such as VEGF, but not of cells. Mice were sacrificed and skinned carefully on day 2. After the implanted chamber was removed from the s.c. fascia, a ring without filters was placed on the same site and photographed. The chamber ring is 10 mm in diameter. The blood vessels newly formed by glioma cells (arrowheads) were morphologically distinguishable from the preexisting background vessels by their zigzagging character.

Close modal
Fig. 6.

The area of dense capillary networks was quantified by an image processor from photographs of the air sac fascia in contact with the chambers containing glioma cells infected with culture medium alone (mock), Ad5CMV-LacZ (lacZ), Ad5CMV-p21 (p21), Ad5CMV-p53 (p53), and Ad5CMVK-p16 (p16) and a chamber containing PBS only (PBS) on day 2. The results are presented as the mean area of capillary networks inside the ring (n = 4). Bars, SD. a, P < 0.05 versus mock, lacZ, and p21. b, P < 0.05 versus mock, lacZ, p21, and p53. c, P < 0.05 versus lacZ. d, P < 0.05 versus mock, lacZ, p21, and p16.

Fig. 6.

The area of dense capillary networks was quantified by an image processor from photographs of the air sac fascia in contact with the chambers containing glioma cells infected with culture medium alone (mock), Ad5CMV-LacZ (lacZ), Ad5CMV-p21 (p21), Ad5CMV-p53 (p53), and Ad5CMVK-p16 (p16) and a chamber containing PBS only (PBS) on day 2. The results are presented as the mean area of capillary networks inside the ring (n = 4). Bars, SD. a, P < 0.05 versus mock, lacZ, and p21. b, P < 0.05 versus mock, lacZ, p21, and p53. c, P < 0.05 versus lacZ. d, P < 0.05 versus mock, lacZ, p21, and p16.

Close modal
Table 1

Effect of Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16 on cell cycle regulationa

U-251 MGU-87 MGSW 1783
36 h72 h36 h72 h36 h72 h
% in G0-G1       
 Ad5CMV-LacZ 53.8 ± 4.9 64.1 ± 4.0 60.6 ± 6.6 68.6 ± 2.9 57.0 ± 3.9 58.8 ± 8.7 
 Ad5CMV-p21 79.8 ± 5.0b 87.6 ± 0.9b 66.7 ± 2.7 71.3 ± 5.0 60.1 ± 3.1 62.6 ± 6.5 
 Ad5CMV-p53 82.1 ± 2.6b 90.7 ± 3.8b 61.2 ± 4.8 70.2 ± 8.1 59.3 ± 2.8 56.1 ± 7.5 
 Ad5CMVK-p16 80.3 ± 4.7b 87.2 ± 9.6b 68.7 ± 3.1 88.4 ± 3.7c 57.8 ± 6.4 56.2 ± 3.8 
U-251 MGU-87 MGSW 1783
36 h72 h36 h72 h36 h72 h
% in G0-G1       
 Ad5CMV-LacZ 53.8 ± 4.9 64.1 ± 4.0 60.6 ± 6.6 68.6 ± 2.9 57.0 ± 3.9 58.8 ± 8.7 
 Ad5CMV-p21 79.8 ± 5.0b 87.6 ± 0.9b 66.7 ± 2.7 71.3 ± 5.0 60.1 ± 3.1 62.6 ± 6.5 
 Ad5CMV-p53 82.1 ± 2.6b 90.7 ± 3.8b 61.2 ± 4.8 70.2 ± 8.1 59.3 ± 2.8 56.1 ± 7.5 
 Ad5CMVK-p16 80.3 ± 4.7b 87.2 ± 9.6b 68.7 ± 3.1 88.4 ± 3.7c 57.8 ± 6.4 56.2 ± 3.8 
a

Each cell line was infected with Ad5CMV-LacZ, Ad5CMV-p21, Ad5CMV-p53, and Ad5CMVK-p16. Cells were analyzed for DNA content using flow cytometry 36 and 72 h after infection. Data are expressed as the mean ± SD of the percentage of G0-G1 (n = 3).

b

P < 0.05 versus Ad5CMV-LacZ.

c

P < 0.05 versus Ad5CMV-LacZ, Ad5CMV-p21, and Ad5CMV-p53.

We sincerely thank Drs. Y. Oka and Y. Sadamoto and K. Oka (Ehime University School of Medicine, Ehime, Japan) for excellent technical assistance.

1
Folkman J. Tumor angiogenesis: therapeutic implications.
N. Engl. J. Med.
,
285
:
1182
-1186,  
1971
.
2
Abdulrauf S. I., Edvardsen K., Ho K. L., Yang X. Y., Rock J. P., Rosenblum M. L. Vascular endothelial growth factor expression and vascular density as prognostic markers of survival in patients with low-grade astrocytoma.
J. Neurosurg.
,
88
:
513
-520,  
1998
.
3
Silbergeld D. L., Ali-Osman F., Winn H. R. Induction of transformational changes in normal endothelial cells by cultured human astrocytoma cells.
J. Neurosurg.
,
75
:
604
-612,  
1991
.
4
Plate K. H., Risau W. Angiogenesis in malignant gliomas.
Glia
,
15
:
339
-347,  
1995
.
5
Ferrara N., Henzel W. J. Pituitary follicular cells secrete a novel heparin-binding growth factor specific for vascular endothelial cells.
Biochem. Biophys. Res. Commun.
,
161
:
851
-858,  
1989
.
6
Senger D. R., Galli S. J., Dvorak A. M., Perruzzi C. A., Harvey V. S., Dvorak H. F. Tumor cells secrete a vascular permeability factor that promotes accumulation of ascites fluid.
Science (Washington DC)
,
219
:
983
-985,  
1983
.
7
Hsu S. C., Volpert O. V., Steck P. A., Mikkelsen T., Polverini P. J., Rao S., Chou P., Bouck N. P. Inhibition of angiogenesis in human glioblastomas by chromosome 10 induction of thrombospondin-1.
Cancer Res.
,
56
:
5684
-5691,  
1996
.
8
Van Meir E. G., Polverini P. J., Chazin V. R., Su Huang H. J., de Tribolet N., Cavenee W. K. Release of an inhibitor of angiogenesis upon induction of wild type p53 expression in glioblastoma cells.
Nat. Genet.
,
8
:
171
-176,  
1994
.
9
Mukhopadhyay D., Tsiokas L., Sukhatme V. P. Wild-type p53 and v-Src exert opposing influences on human vascular endothelial growth factor gene expression.
Cancer Res.
,
55
:
6161
-6165,  
1995
.
10
Bouvet M., Ellis L. M., Nishizaki M., Fujiwara T., Liu W., Bucana C. D., Fang B., Lee J. J., Roth J. A. Adenovirus-mediated wild-type p53 gene transfer down-regulates vascular endothelial growth factor expression and inhibits angiogenesis in human colon cancer.
Cancer Res.
,
58
:
2288
-2292,  
1998
.
11
Fontanini G., Boldrini L., Vignati S., Chine S., Basolo F., Silvestri V., Lucchi M., Mussi A., Angeletti C. A., Bevilacqua G. Bcl2 and p53 regulate vascular endothelial growth factor (VEGF)-mediated angiogenesis in non-small cell lung carcinoma.
Eur. J. Cancer
,
34
:
718
-723,  
1998
.
12
Crew J. P., O’Brien T., Bradburn M., Fuggle S., Bicknell R., Cranston D., Harris A. L. Vascular endothelial growth factor is a predictor of relapse and stage progression in superficial bladder cancer.
Cancer Res.
,
57
:
5281
-5285,  
1997
.
13
Louis D. N. A molecular genetic model of astrocytoma histopathology.
Brain Pathol.
,
7
:
755
-764,  
1997
.
14
Schmidt E. E., Ichimura K., Reifenberger G., Collins V. P. CDKN2 (p16/MTS1) gene deletion or CDK4 amplification occurs in the majority of glioblastomas.
Cancer Res.
,
54
:
6321
-6324,  
1994
.
15
Ueki K., Ono Y., Henson J. W., Efird J. T., von Deimling A., Louis D. N. CDKN2/p16 or RB alterations occur in the majority of glioblastomas and are inversely correlated.
Cancer Res.
,
56
:
150
-153,  
1996
.
16
Jen J., Harper J. W., Bigner S. H., Bigner D. D., Papadopoulos N., Markowitz S., Willson J. K. V., Kinzler K. W., Vogelstein B. Deletion of p16 and p15 genes in brain tumors.
Cancer Res.
,
54
:
6353
-6358,  
1994
.
17
Hamada K., Sakaue M., Alemany R., Zhang W. W., Horio Y., Roth J. A., Mitchell M. F. Adenovirus-mediated transfer of HPV 16 E6/E7 antisense RNA to human cervical cancer cells.
Gynecol. Oncol.
,
63
:
219
-227,  
1996
.
18
Serrano M., Hannon G. J., Beach D. A new regulatory motif in cell-cycle control causing specific inhibition of cyclin D/CDK4.
Nature (Lond.)
,
366
:
704
-707,  
1993
.
19
Fueyo J., Gomez-Manzano C., Yung W. A., Clayman G. L., Liu T. J., Bruner J., Levin V. A., Kyritsis A. P. Adenovirus-mediated p16/CDKN2 gene transfer induces growth arrest and modifies the transformed phenotype of glioma cells.
Oncogene
,
12
:
103
-110,  
1996
.
20
He J., Olson J. J., James C. D. Lack of p16INK4 or retinoblastoma protein (pRb) or amplification-associated overexpression of cdk4 is observed in distinct subsets of malignant glial tumors and cell lines.
Cancer Res.
,
55
:
4833
-4836,  
1995
.
21
Gomez-Manzano C., Fueyo J., Kyritsis A. P., Steck P. A., Roth J. A., McDonnell T. J., Steck K. D., Levin V. A., Yung W. K. A. Adenovirus-mediated transfer of the p53 gene produces rapid and generalized death of human glioma cells via apoptosis.
Cancer Res.
,
56
:
694
-699,  
1996
.
22
Hamada K., Alemany R., Zhang W. W., Hittelman W. N., Lotan R., Roth J. A., Mitchell M. F. Adenovirus-mediated transfer of a wild-type p53 gene and induction of apoptosis in cervical cancer.
Cancer Res.
,
56
:
3047
-3054,  
1996
.
23
Viglietto G., Romano A., Maglione D., Rambaldi M., Paoletti I., Lago C. T., Califano D., Monaco C., Mineo A., Santelli G., Manzo G., Botti G., Chiappetta G., Persico M. G. Neovascularization in human germ cell tumors correlates with a marked increase in the expression of the vascular endothelial growth factor but not the placenta-derived growth factor.
Oncogene
,
13
:
577
-587,  
1996
.
24
Iijima K., Yoshikawa N., Nakamura H. Activation-induced expression of vascular permeability factor by human peripheral T cells: a non-radioisotopic semiquantitative reverse transcription-polymerase chain reaction assay.
J. Immunol. Methods
,
196
:
199
-209,  
1996
.
25
Abe T., Okamura K., Ono M., Kohno K., Mori T., Hori S., Kuwano M. Induction of vascular endothelial tubular morphogenesis by human glioma cells. A model system for tumor angiogenesis.
J. Clin. Investig.
,
92
:
54
-61,  
1993
.
26
Oikawa T., Sasaki M., Inose M., Shimamura M., Kuboki H., Hirano S., Kumagai H., Ishizuka M., Takeuchi T. Effects of cytogenin, a novel microbial product, on embryonic and tumor cell-induced angiogenic responses in vivo.
Anticancer Res.
,
17
:
1881
-1886,  
1997
.
27
Rastinejad F., Polverini P. J., Bouck N. P. Regulation of the activity of a new inhibitor of angiogenesis by a cancer suppressor gene.
Cell
,
56
:
345
-355,  
1989
.
28
Arap W., Nishikawa R., Furnari F. B., Cavenee W. K., Huang H. J. S. Replacement of the p16/CDKN2 gene suppresses human glioma cell growth.
Cancer Res.
,
55
:
1351
-1354,  
1995
.
29
Craig C., Kim M., Ohri E., Wersto R., Katayose D., Li Z., Choi Y. H., Mudahar B., Srivastava S., Seth P., Cowan K. Effects of adenovirus-mediated p16INK4A expression on cell cycle arrest are determined by endogenous p16 and Rb status in human cancer cells.
Oncogene
,
16
:
265
-272,  
1998
.
30
Sandig V., Brand K., Herwig S., Lukas J., Bartek J., Strauss M. Adenovirally transferred p16INK4/CDKN2 and p53 genes cooperate to induce apoptotic tumor cell death.
Nat. Med.
,
3
:
313
-319,  
1997
.
31
Chintala S. K., Fueyo J., Gomez-Manzano C., Venkaiah B., Bjerkvig R., Yung W. K. A., Sawaya R., Kyritsis A. P., Rao J. S. Adenovirus-mediated p16/CDKN2 gene transfer suppresses glioma invasion in vitro.
Oncogene
,
15
:
2049
-2057,  
1997
.
32
Nishimori H., Shiratsuchi T., Urano T., Kimura Y., Kiyono K., Tatsumi K., Yoshida S., Ono M., Kuwano M., Nakamura Y., Tokino T. A novel brain-specific p53-target gene, BAI1, containing thrombospondin type 1 repeats inhibits experimental angiogenesis.
Oncogene
,
15
:
2145
-2150,  
1997
.
33
Walker M. D., Green S. B., Byar D. P., Alexander E., Jr., Batzdorf U., Brooks W. H., Hunt W. E., MacCarty C. S., Mahaley M. S., Jr., Mealey J., Jr., Owens G., Ransohoff J., II, Robertson J. T., Shapiro W. R., Smith K. R., Jr., Wilson C. B., Strike T. A. Randomized comparisons of radiotherapy and nitrosoureas for the treatment of malignant glioma after surgery.
N. Engl. J. Med.
,
303
:
1323
-1329,  
1980
.
34
Russell D. S., Rubinstein L. J. Pathology of Tumours of the Nervous System 5th ed.
219
-247, Edward Arnold London  
1989
.
35
Tanaka T., Cao Y., Folkman J., Fine H. A. Viral vector-targeted antiangiogenic gene therapy utilizing an angiostatin complementary DNA.
Cancer Res.
,
58
:
3362
-3369,  
1998
.
36
Boehm T., Folkman J., Browder T., O’Reilly M. S. Antiangiogenic therapy of experimental cancer does not induce acquired drug resistance.
Nature (Lond.)
,
390
:
404
-407,  
1997
.