RB2/p130 Gene-enhanced Expression Down-Regulates Vascular Endothelial Growth Factor Expression and Inhibits Angiogenesis in Vivo¹

Tumorigenesis is a multistep process that involves several genetic changes resulting in uncontrolled cellular proliferation and inhibition of apoptosis (1). Tumor growth and cellular proliferation are linked by the ability of the tumor to foster proper vascularization from the host to the alien tumor graft. Recent evidence shows that tumors do not grow larger than a few millimeters in size unless vascularized by the host (2). Tumor progression and growth require an appropriate rate of blood vessel formation related to the rate of neoplastic cellular proliferation; otherwise, tumor necrosis and eventual calcification result.

Angiogenesis is driven by a balance between different positive and negative effector molecules influencing the growth rate of capillaries. Various angiogenetic and antiangiogenetic factors have been cloned to date (3, 4, 5, 6, 7). VEGF³and TSP-1 are two of the most well studied. VEGF is a potent tumor-secreted angiogenic factor as opposed to TSP-1, which functions as an antiangiogenic molecule (8, 9). Normal growth results by balanced and coordinated expression of these opposing factors. A switch from normal to uncontrolled vessel growth can occur by up-regulating angiogenesis stimulators or down-regulating angiogenesis inhibitors, suggesting that the angiogenetic process is tightly regulated by the oscillation between these opposing forces(10). The switch to an angiogenic phenotype can occur as a distinct step before progression to a neoplastic phenotype and is linked to epigenetic or genetic changes (11). In support of this theory, mRNA expression of VEGF is up-regulated in aggressive tumor cell lines expressing an activated ras oncogene(12). Conversely, transcription of VEGF is down-regulated in these same tumor cell lines after disruption of the mutant ras allele, thus eliminating VEGF expression and rendering the cells incapable of tumor formation in vivo(13). The switch to an angiogenic phenotype has also been associated with the inactivation of the tumor suppressor gene p53 (14). Conversely, cell lines that are p16 deleted revert to an antiangiogenic phenotype upon the restoration of wild-type cyclin-dependent kinase inhibitor p16 (15).

Different gene therapy approaches using tumor suppressor genes have been tested in vivo to date with varying results(16). Recent studies have indicated that angiogenesis may be regulated to some extent via the p53 tumor suppressor function, but no reports have been published regarding the RB family. The RB gene family includes three members: the Rb tumor suppressor RB/p105, p107, and RB2/p130. These proteins are highly homologous in the “pocket” region, composed of subdomains A and B separated by a spacer region that is highly conserved among each of the proteins (17, 18, 19, 20, 21). This functional domain is targeted by viral oncoproteins and is responsible for many functional interactions(22). Functionally, all of the RB family members show cell type-specific, growth-suppressive properties unique to each member. They each bind and temporally modulate in a distinct manner the activity of specific members of the E2F family of transcription factors and are regulated by phosphorylation in a cell cycle-dependent manner(23). The structural identities of these proteins underlie similar but distinct functional properties. In fact, all three family members inhibit cell cycle progression in the G₁phase of the cell cycle (24, 25, 26). Interestingly, the RB family of proteins exhibits unique growth-suppressive properties that are cell type specific, suggesting that although they may complement each other, their functions are not fully redundant (27).

In several tumor cell lines, pRb2/p130 mediates a G₀-G₁ phase cell cycle arrest, including the human T98G glioblastoma cell line, which is resistant to the suppressive effects of both pRb/p105 and p107(24, 25, 27). Additionally, we have demonstrated that RB2/p130 expression suppresses tumor growth in vivo by inhibiting tumor formation in nude mice as well as causing regression of established tumor grafts using a tetracycline-regulated expression system and retroviral-mediated gene delivery (28, 29). Our results mentioned previously led us to investigate retrospectively the possibility that RB2/p130may modulate tumor progression by affecting the fine-tuned angiogenetic balance, because only tumors of 1–2 mm of diameter can receive all sufficient nutrients by diffusion; therefore, additional growth depends on the development of an adequate blood supply through angiogenesis(30). We decided to test whether a link between RB2/p130 expression and inhibition of angiogenesis may be involved in RB2/p130-mediated tumor suppression/regression.

H23 cells (human lung adenocarcinoma) have been described previously(28). HJC Δ5 cells and their clone HJC 12 (JC-T antigen-transformed hamster glioblastoma) expressing pRb2/p130 under an inducible tetracycline promoter have been described previously(29). Briefly, we used a modified tetracycline-regulated method to create an autoregulatory-inducible RB2/p130 gene expression system created in the HJC 15c cell line, originating from a human polyomavirus-induced (JC virus) hamster brain tumor(29). The parental cell line HJC 15c was used to create the control cell line HJC Δ5 that contains the tetracycline transactivator under the control of the Tetp promoter. HJC Δ5 cells were used to form the HJC 12 cell line, which contains, in addition to tetracycline transactivator, the full length cDNA of the human RB2/p130 gene downstream of the Tetp promoter. In this system, pRb2/p130 expression is repressed in the presence of the antibiotic tetracycline (+) and induced in its absence (−) to 100-fold at the protein level (29). The 293T/17 cell line(human renal carcinoma; Ref. 22) was purchased from the American Type Culture Collection upon authorization of the Rockefeller University. H23 cells were maintained in DMEM supplemented with 10%fetal bovine serum, 2 mml-glutamine. The 293T/17 cell line was maintained in DMEM supplemented with 10% heat-inactivated fetal bovine serum and 2 mm l-glutamine. HJC Δ5 and HJC 12 cells were grown in DMEM supplemented with 5% FCS (Sigma Chemical Co., St. Louis, MO) and the antibiotics streptomycin (10 mg/ml) and penicillin (100 units/ml) and in the presence or not (+/−)of 2 μg/ml tetracycline (Sigma).

Retroviral and adenoviral vectors expressing RB2/p130 or controls expressing the bacterial β-Gal (Lac-Z) or the puromycin resistance (Pac) gene alone have been described previously (28, 31). Briefly, retrovirus-mediated gene transfer studies were carried out with a MLV-based system(28). Transient and DNA cotransfection of the 293T/17 cells using PHIT60 (CMV-MLV-gag-pol-SV40 ori) and PHIT456(CMV-MLV-amphotropic env-SV40 ori) vectors, along with MSCV-based transfer vectors MSCV-Pac, MSCV-Pac-LacZ, and MSCV-Pac-RB2/p130, were performed by calcium phosphate precipitation (28). The retroviral supernatant was collected 48 h after transfection, filtered through 0.45 μm filters, and titered as described previously (28) to produce retroviruses carrying the puromycin resistance gene alone or in combination with the Lac-Z gene or the RB2/p130ORF, respectively. Viral titers of 1 × 10⁷ infectious units/ml were obtained(28).

Adenoviruses were generated by subcloning the full-length ORF of the RB2/p130 gene into the pAd.CMV-Link1 vector to form the Ad.CMV-RB2/p130 virus, as described previously(31). The pAd.CMV-Link1 vector alone (to produce the Ad-CMV virus) was used as a negative control to assay the effects of viral infection alone without delivering a transgene. The above-mentioned viruses were generated by cotransfection of the constructs mentioned previously with an adenoviral backbone into the packaging cell line 293 primary embryonal human kidney cells,transformed by sheared human adenovirus type 5. The adenoviruses were recovered, screened, and expanded as described previously(31). After purification by sequential equilibrium,density gradients using CsCl viral stocks were made at 5 × 10¹² particles/ml and stored at −80°C in a solution containing 10% glycerol. A viral titer of 22 × 10⁹ pfu/ml was determined by plaque assay for the Ad-CMV and Ad-CMV-RB2/p130 viruses. Infection of nonpermissive cells confirmed that the viruses were replication defective.

H23 cells were grown to 70% confluency then infected with 50 multiplicity of infection of adenoviruses carrying RB2/p130ORF or with the control Ad-CMV. After 14 h, the medium was changed, and the cells were harvested after a total of 48 h after infection.

VEGF Northern blot analysis was performed essentially as described previously (32). Briefly, RNA was extracted using the RNAzol kit (Tel-Test, Inc., Friendswood, TX), following the manufacturer’s instructions. The RNA was resolved on a 1% agarose gel containing 6.6 m formaldehyde, transferred to a Zeta Probe(Bio-Rad, Hercules, CA) membrane, and hybridized at 65°C with a ³²P-labeled cDNA probe containing either the 200-bp fragment of the human VEGF sequence common to all four known isoforms of VPF/VEGF protein or TSP-1 (32, 33). The amount of RNA loaded in each lane was evaluated by ethidium bromide gel staining of the gel before the transfer. The TSP-1 probe was purchased from the American Type Culture Collection.

Rabbit polyclonal anti-VEGF was the kind gift of Genentech, Inc. (San Francisco, CA). Purified antimouse CD31 (PECAM-1), clone MEC 13.3 was purchased from (PharMingen, San Diego, CA). Anti-VEGF was used at a dilution 1:500, and anti-CD31 was used at a dilution of 1:50, following the manufacturer’s instructions for immunohistochemical analysis.

VEGF staining intensity was graded on a scale of 0 to 3: 0, no detectable staining; 1, traces of staining; 2, moderate amount of diffuse staining; and 3, a large amount of diffuse staining. This grading scale is a modification of that of Takahashi et al.(34).

Intratumoral microvessels were highlighted by immunostaining different serial, formalin-fixed, paraffin-embedded sections of the same tumor graft with anti-CD31. IMD was determined as described previously(35). Briefly, CD31-stained sections underwent an individual microvessel count on a ×400 magnification in the areas of most intense neovascularization (hot spots). IMD was expressed as microvessels/mm².

The VEGF ELISA was performed as described previously using the anti-VEGF from Genentech, Inc. (1:2000 dilution) and an antirabbit horseradish peroxidase-conjugated antibody (Amersham, Arlington Heights, IL; 1:5000 dilution) as secondary antibody and the 3,3′,5′5-tetramethylbenzidine liquid substrate system following the manufacturer’s recommendations (Sigma; Ref. 36).

Luciferase assay was performed by transfecting a total of 3 μg of either mouse VEGF promoter (37) or an artificial E2F promoter containing three consecutive E2F consensus binding sites(38) linked to the luciferase reporter gene for each point in the HJC 12 cells, in the presence of the antibiotic tetracycline (+;uninduced status) and in its absence (−; induced pRb2/p130 protein status). HJC 12 cells were plated at 60% confluency in six-well dishes the day before the experiment, and transfections were performed by the standard calcium-phosphate method as reported previously(29). Normalization was performed by cotransfecting a total of 1 μg of CMV Lac-Z (Promega, CA) for each experimental point. The experiment was performed in triplicates and repeated twice. Luciferase activity was assayed using the luciferase kit assay according to the manufacturer’s instructions (Promega Corp.,Madison, WI) and measured using a luminometer (Corning Costar Corp.,Cambridge, MA).

Protein concentration was assayed by Bradford analysis (Bio-Rad Laboratories, Inc., Melville, New York) and confirmed by running 10μg of protein on a 12% SDS-polyacrylamide gel (SDS-PAGE); staining was with Coomassie blue. For Western blotting purposes, an equal amount of 100 μg of protein extract for each sample was electrophoresed into 12% SDS-polyacrylamide gels (SDS-PAGE) and transferred to 0.2 μm nitrocellulose membranes (Schleicher & Schuell, Germany). The loading and transfer of equal amounts of protein were confirmed by staining the membranes with Red Ponceau (Sigma). Membranes were quenched at 4°C overnight in a solution of TBS-T (Tris-buffered saline + 0.5% Tween 20) and 5% dry milk for blocking nonspecific binding. Either primary rabbit polyclonal anti-VEGF (Genentech) or anti-VEGF(Santa Cruz Biotechnology, Inc., Santa Cruz, CA) diluted 1:200 in a solution of TBS-T and 3% dry milk was used independently to incubate the blots. After several washes in a solution of TBS-T, the blots were incubated with a solution of TBS-T containing an antirabbit secondary antibody (horseradish peroxidase conjugated; Amersham, Life Science),diluted 1:20,000 for 1 h at room temperature. The blots were then washed several times in TBS-T, reacted with a ECL (Enhanced Chemiluminescence kit; DuPont NEN, Boston, MA), and exposed to X-ray films.

Animal care and humane use and treatment of mice were in strict compliance with the following: (a) institutional guidelines;(b) the Guide for the Care and Use of Laboratory Animals(National Academy of Sciences, 1996); and (c) the Association for Assessment and Accreditation of Laboratory Animal Care International. Tumors were generated by the s.c. injection of 2.5 × 10⁶ H23 or of 5 × 10⁶ HJC Δ5 or HJC 12 cells into nudemice (female nu/nu-nuBR outbred,isolator-maintained mice, 4–5 weeks of age, from Charles Rivers Wilmington, MA), as described previously (28, 29).

For H23 injected cells, when the tumors reached a volume of ∼20 mm³ after 15 days, each tumor was transduced with 5 × 10⁶ retroviruses carrying the Pac gene alone or the Pac gene and the Escherichia coli β-Gal (Lac-Z) gene as control or the Pac gene and RB2/p130 ORF with three animals/group by direct injection of 20 μl of retroviral supernatant directly into each of the tumors.

For the HJC nude mice group, the mice were treated with tetracycline for 4 days prior to injection. The mice were injected s.c. along their left and right flanks at two sites/mouse with 5 × 10⁶ cells/flank while under anesthesia with isopropane gas. There were four groups of animals with three animals/group. Two groups were injected with HJC 12 cells, and treatment with tetracycline continued after injection in one group(12+), whereas another group (12−) ceased to be administered tetracycline after injection of the cells. The two control groups were injected with HJC Δ5 cells, and one (Δ5+) continued to receive tetracycline while the other control group (Δ5−) did not.

Animals were sacrificed by CO₂ asphyxiation when Pac and Lac-Z retrovirus-transduced tumors or HJCΔ5 (± tetracycline) and HJC 12 tumors (+ tetracycline) reached a size of 300–350 mm³. Tissues to be sectioned were placed in OTC (Sakura Finetek USA, Inc., Torrance, CA), frozen in liquid nitrogen, and stored at −80°C or preserved in neutral-buffered formalin at 4°C before embedding in paraffin.

We have shown previously that RB2/p130 enhanced expression inhibits tumor formation (28, 29) and induces tumor regression (28, 29). Interestingly, we found that some of the tumors treated with retroviruses delivering RB2/p130underwent central necrosis and subsequent calcification. Following these observations, we tested the hypothesis that RB2/p130overexpression could inhibit angiogenesis. We chose to study the expression levels of two well-studied proteins involved in angiogenesis, VEGF and TSP-1.

H23 cells were infected with adenoviruses carrying RB2/p130or with the control Adeno-CMV and harvested after 48 h. Interestingly, Northern blot analysis with a probe against VEGF showed a down-regulation of 2- fold of the vascular endothelial growth factor upon overexpression of RB2/p130 (Fig. 1). On the other hand, no modification of TSP-1 was observed in the same experimental conditions (Fig. 1). Because low abundance of gene expression can result from either enhanced mRNA degradation or promoter regulation, we thought to analyze the effects of forced RB2/p130 gene expression on the VEGF promoter. HJC 12 cells transfected with the VEGF promoter and cultured in the absence of the antibiotic tetracycline (RB2/p130 induced) showed 2–3-fold down-regulation with respect to the uninduced HJC 12 (+ Tet) cells and to the vector control-transfected cells in either the induced (−Tet)or uninduced status (+ Tet; Fig. 2). A promoter containing E2F consensus binding sites linked to the luciferase reporter gene was used as a positive control for pRb2/p130 transcriptional repression activity. These results led us to explore the VEGF protein abundance in vitro and in vivoafter enhanced RB2/p130 expression.

To test the hypothesis that RB2/p130 may modulate VEGF protein expression in vitro, we studied (39)the conditioned culture medium of H23 cells transiently transduced with either adenovirus carrying RB2/p130 or with the CMV control adenovirus, by means of an ELISA. Additionally, we tested the conditioned medium of the HJC 12 cells in which the RB2/p130expression is regulated by a tetracycline-inducible promoter(29). In both systems, overexpression of RB2/p130 resulted in a down-regulation of the VEGF protein abundance by 3-fold with respect to the controls (Fig. 3). To further our analysis, we also tested VEGF abundance in the intracellular compartment. H23 cells were transiently transduced with either adenovirus carrying RB2/p130 or with the CMV control adenovirus. We also tested protein extracts from HJC 12 cells in which the RB2/p130 expression is regulated by a tetracycline-inducible promoter. Western Blot analysis using rabbit polyclonal antibodies against VEGF showed a 2–3-fold reduction of intracellular protein abundance upon enhanced RB2/p130expression (Fig. 4).

Using samples from two previous studies in which we showed that RB2/p130 inhibits tumor formation and induces tumor regression in nude mice (28, 29), we were able to identify a novel mechanism of tumor inhibition for RB2/p130. Serial sections of tumors grown in nudemice and treated or untreated with RB2/p130 were immunostained for VEGF and CD31. CD31 is a specific marker for endothelial cells (40). We chose to grade the VEGF staining on a scale from 0 to 3, following previous work by Takahashi et al. (34), with some modification. We considered a score of 0 equal to no detectable staining; 1, traces of staining; 2, a moderate amount of diffuse staining; and 3, a large amount of diffuse staining. Interestingly, RB2/p130overexpression caused VEGF immunostaining to drop from a large amount of diffuse staining (score, 3), characteristic of the control samples(Fig. 5, A–C, and Fig. 6, A–D), to traces of staining (score, 0; Fig. 5,Dand Fig. 6, E and F) in both the two tumor graft groups examined (see also Table 1).

IMD assessment showed at least an 81% (confidence interval,1.95–10.5) reduction of microvessels count after CD31 immunostaining in all tumor grafts (H23 and HJC) in which RB2/p130 was overexpressed (Table 1). Fig. 5, F–H, shows a few representative examples of microvessel density in the HJC Δ5 (+ Tet),HJC Δ5 (− Tet) and HJC 12 (+ Tet) control tumor grafts,respectively. Fig. 6, G and H, instead show samples of H23 tumor grafts treated with the control retroviruses carrying Pac or β-Gal, respectively. Fig. 5,I, however, demonstrates very poor microvessel density upon induction of RB2/p130 expression in HJC 12 (− Tet)tumor grafts as evidenced by CD31 immunostaining. Fig. 6,I at low magnification power (×100) shows poor microvessel density in a H23 tumor graft treated with retroviruses carrying RB2/p130. Additionally, Fig. 6,I contains on its upper side a portion of normal nude mouse tissue demonstrating a normal neurovascular formation that was stained by the CD31 antibody, proving that the lack of CD31 staining is indeed specific to enhanced pRb2/p130 expression in tumor tissues. Fig. 6,J is a higher magnification field (×400) of Fig. 6,I, showing the only vascular formation present on this particular slide. Finally, Figs. 5,J and 6 K show the specificity of the CD31 staining to neurovascular bundles in normal embryonal mouse lung endothelium in the conditions used. The effects upon VEGF staining intensity and IMD were specific to pRb2/p130 expression because withdrawal of tetracycline from the HJC Δ5 tumors did not alter VEGF intensity and actually enhanced IMD.

We have shown previously that RB2/p130 potently inhibits tumor formation in nude mice and causes tumor regression of fully established tumor grafts (28, 29). In the present retrospective study, RB2/p130 significantly decreased VEGF RNA and protein expression in vitro and in vivo in two different cell types, glioblastoma and lung adenocarcinoma, in both rodent and human tumor cell lines, using either retrovirus/adenovirus-mediated gene transfer or a tetracycline-regulated gene expression system. Additionally, enhanced RB2/p130 gene expression down-regulated the activity of the VEGF promoter in a tetracycline-regulated pRb2/p130 system. We analyzed the VEGF promoter for putative E2F regulatory binding sites and found that it does not contain any known responsive site of regulation of the RB pathway. Therefore, the down-regulation of the VEGF promoter that we have observed could be attributable to an indirect mechanism that needs to be further investigated. The fact that these data were reproduced using different expression systems indicate also that the VEGF down-regulation observed at the RNA, promoter, and protein levels was not attributable to a mere viral bystander effect but was contingent upon the enhanced expression of the RB2/p130 gene within the tumor itself. Previous reports have indicated that the tumor suppressor gene p53 and the cyclin-dependent kinase inhibitor p16 control tumor angiogenesis by regulating TSP-1 or VEGF expression(9, 15). However, the exact mechanism by which p53 and p16 operate the VEGF down-regulation is still unknown. VEGF and TSP-1 are two important factors regulating angiogenesis and antiangiogenesis,respectively. High tumorigenic potential is associated with elevated levels of VEGF (32). On the other side, TSP-1 overexpression has been reported to suppress tumor growth and metastasis potential in some cell types (41). Additionally, another group showed instead that wild-type p53 suppresses VEGF expression in an anaplastic thyroid carcinoma cell line(42). Our data that forced expression of RB2/p130 significantly determined tumor regression and inhibited VEGF RNA abundance by 2-fold and protein expression as well as VEGF promoter activity by 3-fold, causing at least 81% (confidence interval, 1.95–10.5) reduction in vessel formation, indicate a novel tumor-regulatory property for the RB-related gene RB2/p130. This new growth regulatory feature seems to involve the control of the basic nutritional supplies that the tumor extracts from the host. Moreover, the knowledge that p53, p16, and now also pRb2/p130 regulate angiogenesis via inhibition of the vascular endothelial growth factor,suggests that cell cycle regulatory proteins generally acting in the G₁ phase of the cell cycle could have similar effects, or that these proteins control distinct pathways with a common end point. This similarity, along with the fact that DNA tumor viruses simultaneously evolved the ability to repress both p53 and RB family function to accomplish cellular transformation, suggests a cooperation between the proteins in their strategies to regulate proliferation and tumor progression.

The roles of these effectors in tumorigenesis need to be studied more closely to define the mechanisms behind their inhibition of angiogenesis. This would aid in designing more targeted and effective combined gene therapy strategies.

We thank Dr. K. Hubner (Thomas Jefferson University,Philadelphia, PA) for providing the H23 cell line and the normal mouse tissues used to test the CD31 and the VEGF antibodies and Dr. P. B. Fisher (Columbia University, College of Physicians and Surgeons, New York, NY) for providing the mouse VEGF promoter.