Deletion of Yin Yang 1 Protein in Osteosarcoma Cells on Cell Invasion and CXCR4/Angiogenesis and Metastasis

Osteosarcoma is the second leading cause of cancer-related deaths in the pediatric age group and the most common primary bone tumor in general. Tumors are often localized to the distal femur and proximal tibia region and the survival rate is between 50% and 65% (1). Although current patient management practices have significantly improved surgical resection and long-term outcome, 25%/50% of patients with initial diffusion succumb to lung metastasis (2–4). Indeed, systemic spread of this tumor already occurring at the time of detection and its genetic variability contribute to make an adequate eradication of this tumor with the current treatment regimens difficult (4). It seems from the most recent studies and clinical trials that a multitargeted approach to therapy is necessary (4). However, the rate of mortality for osteosarcoma is basically the same as that some decades ago (1–4). Thus, we need to understand more the molecular mechanisms involved in this type of cancer.

Recently, we have provided the first evidence that the Yin Yang 1 protein (YY1) is overexpressed in osteosarcoma tumors and is correlated in a positive manner with their more aggressive phenotypes (5, 6). However, these observations are not informative on whether YY1 plays a key role in the development of such an aggressive undifferentiated phenotype. Because metastatic implantation may induce a general perturbation of cell cycle and angiogenesis (7, 8), we also decided to investigate this issue. Besides, chemokines and chemokine receptors (CXCR) play a central role in determining the organ target of the tumor cells (9, 10) as well as the blood supply via angiogenesis (11, 12). Indeed, recent studies have indicated that CXCR4 plays a role in regulating metastasis of many solid tumors and their neovasculature (13–17). More relevant is the fact that CXCR4 may interact with YY1 (13) and is expressed in osteosarcoma cells and in mouse models of metastasis (14–17). Therefore, there is a strong rationale behind investigating the role of this complex network involving YY1, CXCR4, angiogenesis, and the vascular endothelial growth factor (VEGF) in the pathogenesis of osteosarcoma and its metastatic dissemination.

Here, we address the effects of YY1 silencing on osteosarcoma cells and metastatic implantation and angiogenesis by experiments both in vitro and in vivo.

Osteosarcoma cell lines and transfection. The SaOs-2 cell lines were grown in DMEM (Life Technologies) plus 10% fetal bovine serum (FBS), penicillin/streptomycin, and glutamine as described previously (5, 6). Stable transfection with small interfering RNA (siRNA) YY1 plasmid and empty vector was obtained using the Lipofectamine 2000 (Stratagene) protocol. Cells were selected in medium containing 100 μg/mL G418 (Sigma-Aldrich). The rescue clone was produced by cotransfecting siRNAYY1 plasmid together with pCMV tag mycYY1 plasmid (6).

Interference construct. Selection of YY1 interference oligonucleotides was performed using the sequences available online.⁷

Immunofluorescence. Cells were grown on coverslips and transiently or stably transfected as above and then fixed in paraformaldehyde for 5 min and stained with an anti–green fluorescent protein (GFP; 1:300). For myc tag staining, 10⁵ living cells were incubated in 10 μL of cold PBS in the presence of tag myc (1:100) antibodies (Invitrogen) for 30 min and then washed twice with PBS and incubated with secondary antibodies, as described above. The secondary antibodies used were anti-rat Alexis 594 and anti-mouse Alexis 488 (1:800; Molecular Probes). The preparation was viewed in a Leica confocal TCSS82 microscope and in a Zeiss microscope Axiophoto.

Western blot and densitometric analysis. Total protein extracts (cells or tissues) were separated by 10% SDS-PAGE, transblotted onto a nitrocellulose membrane (6, 18–20), and incubated with the following antibodies: YY1 and CXCR4/fusin from Santa Cruz Biotechnology, CXCR4 mouse monoclonal antibody clone and VEGF and VEGF receptor 3 (VEGFR3) monoclonal antibodies from R&D Systems, goat polyclonal metastasis tumor antigen (MTA) 1 and goat polyclonal MTA2 from Santa Cruz Biotechnology (1:1,000 for 1 h at room temperature), and γ-tubulin protein T6557 from Sigma (1:10,000 for 1 h at room temperature). Semiquantitative assessment of band intensity was made by Image-Pro Plus (Media Cybernetics).

Proliferation and adhesion assay. SaOs-2 and siRNAYY1-SaOs-2 clones and 20,000 cells were seeded in 24-well plates in medium and numbered each day after seeding using trypan blue dye exclusion and a hemocytometer chamber. The adhesive ability was analyzed using the Cytomatrix cell adhesion strips coated with human collagen type IV, fibronectin, and laminin (Chemicon International). Briefly, 40,000 cells per well (96 multiwells) were seeded and incubated for 1 h at 37°C. Adherent cells were fixed and stained with 0.2% crystal violet/10% ethanol and read at 485 nm on a microplate reader. All the experiments were performed in triplicate.

Boyden chamber invasion assay. Invasion assays were performed with Boyden chambers with 8-μm filters (BD Biosciences) coated with a 1:4 dilution of Matrigel (BD Biosciences). Cells (5 × 10⁴) were seeded in serum-free medium containing 0.1% bovine serum albumin (BSA) in the upper compartment; DMEM plus 10% FBS was placed in the lower compartment of the chamber as a chemoattractant. For chemoinvasion experiments, 5 × 10⁴ cells were seeded in 100 μL of serum-free medium containing 0.1% BSA in the upper compartment; recombinant stromal-derived factor-1 (SDF-1; CXCL12; 100 ng/mL; PeproTech) was either added or not to the bottom chambers. For neutralization experiments, cells were preincubated for 1 h with T22 and ALA scrambled peptides as described previously (Primm; ref. 17) at a concentration of 1 to 10 μg/mL or CXCR4 antibodies at 50 μg/mL and then SDF-1 was added or not to the bottom chambers. The Boyden chambers were incubated for 72 h at 37°C and the cells present in the lower chamber were counted. All the experiments were repeated six times.

Mice treatment and magnetic nuclear resonance protocol of metastatic monitoring. Five-week-old CD1 female nude mice obtained from an outbreed background were purchased from Charles River Laboratory. The housing and handling of mice were in accordance with institutional guidelines and compliant with national (Ministero della Salute, Rome, Italy) and international laws (European Community and National Institutes of Heath, Bethesda, MD) and policies (Dr. C. Arra, who is the head of the preclinical unit at Pascale Foundation, Naples, Italy). Quality standards of laboratories at the Pascale Foundation and II University of Naples (Italy) are in accordance with rules established by the Italian Ministry of Health and the European College of Laboratory Animal Medicine, whereas quality standards of laboratories at the University of California at Los Angeles (United States) are in accordance with the standards of the Association for Assessment and Accreditation of Laboratory Animal Care. All experimental procedures were approved by the local ethical committees on preclinical studies. The mice were allowed to acclimate to their new environment for 1 week after the arrival. They were housed in disinfected polycarbonate mouse cages, maintained in cabinets with laminar flow at 28°C under controlled artificial lighting (12-h light/12-h dark), and given ad libitum access to rodent chow and water during the study. We performed the experiments on two groups of 10 animals. The mice were injected through the tail vein with 1 × 10⁶ cells. Those of the first group received SaOs-2 cells (in 100 μL PBS), whereas those of the second group were injected with siRNAYY1-SaOs-2 cells (clone 9). Cell viability was determined by trypan blue dye exclusion. After 1 week, the cell injection was repeated. Each group was then subdivided into control and treated mice. The mice that were treated received via i.p. injection 5 mg/d/mouse of T22 peptide, whereas the control group was injected with ALA scrambled peptides (Primm; ref. 17) in 100 μL of sterile PBS from day 10 to day 35 after cell injection. The mice were euthanized using CO₂ asphyxiation after a mean of 13 weeks. During this period, to follow the development of lung metastasis (and lymph node alterations), the animals were subjected to serial magnetic resonance imaging (MRI) analysis with 1.5 Tesla (MAGNETOM Symphony, syngo MR 2002B, Siemens). The MRI sequences used were the following: a precontrast T1-weighted, two-dimensional, turbo spin echo sequence [repetition time/echo time (TR/TE) of 400/13 ms, 2-mm slice thickness, 10 mm gap, three signal averages, 205 × 256 matrix, 10 cm field of view, 4 min acquisition time]; a T2-weighted, two-dimensional, turbo spin echo sequence (TR/TE of 4,000/95 ms, 2-mm slice thickness, 10 mm gap, three signal averages, 205 × 256 matrix, 10 cm field of view, 4.47 min acquisition time); and a postgadolinium-enhanced (200 μL Magnevist, Schering) T1-weighted, two-dimensional, turbo spin echo sequence. The T1-weighted turbo sequences, precontrast and postcontrast, were obtained in the sagittal and coronal planes and the T2-weighted turbo sequence was obtained in the coronal plane. The total scan time was ∼15 min.

Angiogenesis by direct in vivo assay. For direct angiogenesis assay, the direct in vivo assay (DIVA) kit was used (Trevigen). Briefly, angioreactors were filled with Matrigel with or without the angiogenic factor [fibroblast growth factor-2 (FGF-2)]. They were incubated at 37°C for 1 h to allow gel formation. At day 30 after cell injection, 1.5 cm incisions were made on the mice skin, and the angioreactors were implanted into the dorsal flank. The incisions were sutured and overlayed with an antiseptic solution. In each mice, two angioreactors were implanted: the whole angioreactor in the left flank and the positive control (angioreactor coated with FGF-2) in the right. At the end of the experiment, the angioreactors were collected and the new vessel formation was determined by FITC-lectin staining. After staining, the vessels were washed and the fluorescence was measured in 96 multiwell plates using an HP spectrofluorimeter model (excitation, 485 nm; emission, 510 nm; Perkin-Elmer). The mean relative fluorescence ± SD for five replicate assays was determined.

Histology and immunohistochemistry. Vessels from angioreactors were fixed in 10% formalin. Computer-assisted imaging analysis of histologic sections, which were prepared and stained with H&E methods, was carried on as described previously (6, 12, 21). The harvested lung metastases were fixed and embedded in paraffin for histologic examination according to standard conditions. Specimens of 4 μm were stained with H&E (6, 12, 21, 22) and with human monoclonal antibodies for YY1; polyclonal CXCR4 antibody and monoclonal antibodies for MTA1 and MTA2; and VEGF, VEGFR3, CD31, and CD34 monoclonal antibodies (see above). Serial sections were examined with a Zeiss Axiophoto microscope (Cal Zeiss, Inc.) at ×400 magnification and fluorescence images were acquired and analyzed with the Evolution VF fast digital camera (Media Cybernetics; 6, 12, 21, 22).

Soft agar. Cell suspensions (10,000–30,000 in 60-mm dish) were plated in semisolid medium (0.3% agar in DMEM plus 10% FBS) and incubated at 37°C in a humidified atmosphere with 5% CO₂. Colonies were counted after 10 to 14 days.

Cell cycle analysis. Cells were fixed with 70% ethanol after two washes with PBS, stained with 20 μg/mL of propidium iodide, and analyzed by flow cytometer. For flow cytometry analysis, 5 × 10⁵ cells were incubated with phycoerythrin (PE)-CXCR4 (FAB107P; R&D Systems) for 30 min at 4°C in 0.1% BSA/PBS, and after washing with PBS, the samples were analyzed using FACSCalibur (Becton Dickinson).

RNA extraction, microarray analysis, and reverse transcription-PCR. We have a long trust experience with genomics and microarray analysis (reviewed in refs. 23, 24). Total RNAs were extracted using Trizol solution (Invitrogen) according to the manufacturer's instructions. Briefly, 5 μg of total RNAs from a pool of siRNAYY1-SaOs-2 clones were used to obtain labeled cDNA. The first- and second-strand cDNA synthesis was performed using the GeneChip One-Cycle cDNA Synthesis kit (Affymetrix), as described previously in detail (23, 24). Labeled cRNA was prepared using the GeneChip IVT Labeling kit (Affymetrix) according to the manufacturer's instructions. The cRNA was used for hybridization onto the Affymetrix Human Genome 2.0 probe array cartridge. The probe arrays were scanned at 560 nm using a confocal laser scanning microscope (GeneChip Scanner 3000). The readings from the quantitative scanning were analyzed by the Affymetrix gene expression analysis software (19). Total RNA (1 μg) was reverse transcribed with SuperScript III (Invitrogen). Results were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH), which was used as a housekeeping gene. The intensity of the amplified bands was quantified by densitometry and referred to that obtained with GAPDH (Quantity One, Bio-Rad). Finally, matrix metalloproteinase (MMP) mRNA was evaluated by reverse transcription-PCR (RT-PCR) following the protocol described previously (25). The amount of specific mRNA bands was calculated by densitometry and reported as ratio (arbitrary units) in comparison with those of β-actin. Primer sequences and corresponding PCR conditions are shown in Supplementary Table S1.

Statistical analysis. All histologic and immunohistochemistry data were analyzed blindly by two independent pathologists. Data were expressed as mean ± SD. Differences among means were analyzed using Student's t test. Fisher's exact test was used for frequency data. Significance was established at P < 0.05.

Selection and characterization of siRNAYY1 clones in SaOs-2 cells. We have transfected SaOs-2 cells with siRNAYY1 plasmid. Stable transfectant clones were selected in G418 (200 μg/mL), and the reduction of YY1 protein expression was evaluated by Western blot (Fig. 1A,, a). Thus, we have selected four clones showing a 60% to 70% YY1 reduction compared with SaOs-2 cells (Fig. 1A,, a). We were never able to select siRNAYY1 SaOs-2 clones with completed ablation of YY1. The rescue clone was produced by cotransfecting pBLOCK-iT YY1 plasmid with pCMVmycYY1 (6) and selected in G418. The expression of both constructs was evaluated by immunofluorescence using specific antibodies and Western blot (Fig. 1A , a and b, left).

siRNAYY1 reduces the growth and adhesion ability in vitro. The siRNAYY1 SaOs-2 clones (clones 1, 8, 9, and 10) grew more slowly in monolayer compared with the control (SaOs-2 empty vector and rescue clone; Fig. 1B,, a). This effect was impressive in anchorage-independent conditions (soft agar), where siRNAYY1 clone 9 did not form colonies (Fig. 1A,, b). SaOs-2 cells had the highest rate of cell cycling, but this was not significantly greater than that for siRNA clones (P < 0.05; Fig. 1B,, b). In collagen IV, cell adhesion assays (see Fig. 1C,, a), clones 1, 8, 9, and 10 were less adherent (2-fold) than controls (P < 0.01 versus SaOs-2). Lower differences were observed in fibronectin I adhesion assays (Fig. 1C , b), whereas no differences were detected on laminin assay (data not shown). The collagen IV and fibronectin mRNAs were measured by RT-PCR.

The silencing of YY1 alters the cell invasion in noncoated and coated Matrigel filters. We have evaluated the effect of YY1 silencing on cell migration and invasion in the presence or absence of SDF-1. YY1 silencing increased chemotaxis stimulated by SDF-1 (2-fold higher than SaOs-2; P < 0.001; Fig. 2A,, a). Whereas in Matrigel-coated filter invasion assay the YY1 silencing reduced (4-fold; P < 0.001) cell migration in response to SDF-1, compared with controls (P < 0.001; Fig. 2A,, b) and in the presence of 100 ng/mL SDF-1 (CXCR4 ligands) as chemoattractant, the clones showed a 4-fold lower cell migration (P < 0.001). The invasion activity was specifically reduced with the addition of T22 blocking peptide (100 ng/mL; Fig. 2A,, b). Similar data were obtained with the CXCR4 antibody (50 μg/mL; data not shown). Moreover, we have investigated whether the observed differences were a consequence of reduced Matrigel degradation activity or of different CXCR4 expression. As shown in Fig. 2B (a), the YY1-silenced clone 9 expressed a lower level of MMP2 compared with SaOs-2 cells (Fig. 2C). These data indicate that the silencing of YY1 down-regulates MMP transcription, causing a reduction in cell migration trough a Matrigel-coated filters. Protein and fluorescence-activated cell sorting (FACS) analysis (Fig. 2C , a and b) indicated that CXCR4 protein was a little higher in clones than in control cells, whereas the mRNAs were down-regulated. YY1 was positively correlated only with CXCR4 mRNA, whereas increased cell surface expression of CXCR4 may be the result of altered regulation independent of the transcription/translation mechanism. Ubiquitination of CXCR4 is a modification regulating the expression of CXCR4 posttranslationally found deregulated in tumor cells (11).

Role of YY1 in orthotopic murine model of metastasis. One group of 10 nude mice was injected twice with 1 × 10⁶ SaOs-2 cells in the tail vein. Another group of 10 mice received 1 × 10⁶ siRNAYY1 clone 9 cells. On the basis of our and other (13–17) preliminary experiments, five mice of each group received 50 μg/d of T22 peptide for 30 days (starting after 2 weeks from cell infusion). The progressive development of metastases was monitored by nuclear magnetic resonance (NMR). After a mean of 13 weeks, NMR revealed the presence of big lymph nodes and an opaque area in mice's lungs (Fig. 3A , a); the mice were then sacrificed and the lungs were analyzed (see below).

Systemic angiogenesis. In parallel experiments, to have a picture of angiogenesis in vivo, during the metastatic implantation of SaOs-2 and clone 9 cells, two DIVA angioreactors per each mouse were implanted into the dorsal flank (Fig. 3A,, b). Angioreactors were collected at the end of the experiment and the new vessels were quantified. As shown in Fig. 3A (b), the angioreactors contained several newly formed vessels near the open end that invaded the Matrigel to form a vasculature structure with circulating RBCs. The histologic examination of the vessels revealed the presence of vasculature formation and more lumen in SaOs-2 cell–bearing mice compared with clone 9 mice (Fig. 3B,, a). The direct quantification of endothelial cell presence was obtained by incubating the vessels with FITC-lectin and measuring their fluorescence. The assay of vessels from SaOs-2–inoculated mice revealed a 50% increase in FITC-lectin–positive cells (P < 0.05 versus SaOs-2) compared with clone 9 mice (Fig. 3B,, b), indicating that the silencing of YY1 interfered directly with angiogenesis in vivo. In addition, in this case, the injection of T22 peptide reduced the newly formed vessels (−40%) in SaOs-2–bearing mice (P < 0.01 versus SaOs-2). T22 was ineffective in preventing vessel formation in clone 9 cell–bearing mice (Fig. 3B , b), suggesting that YY1 acts downstream of CXCR4 and cross-interacts with CXCR4/angiogenesis pathway.

Examination of lung metastasis. A preliminary evaluation by PCR confirmed the expression of bone alkaline phosphatase in all metastases and cells (Fig. 3B,, b). Detailed and blinded bioptic examination revealed that nine mice (n = 10) inoculated with SaOs-2 cells developed metastasis, whereas only four mice from the clone 9 group did (n = 10; as shown in Supplementary Table S2). Macrometastasis was present only in SaOs-2–bearing mice and their computerizing mean area (expressed in μm²) was 10-fold bigger than the area of the metastases in mice inoculated with clone 9 cells (see Supplementary Table S2 and Fig. 4A,, a). Relevantly, the administration of T22 peptide reduced the metastasis size (10-fold lower) in SaOs-2–bearing mice, whereas it was not significant for clone 9 (Supplementary Table S2). Serial histologic examination confirmed that SaOs-2–induced metastasis was bigger than that of clone 9 (Fig. 4A,, a compared with B, a). Because literature data did not histologically characterize this kind of metastasis very well, we have handled carefully this issue. Here, we report the presence of osteoid matrix deposition in core metastasis, as indicated by arrows in the red area of Fig. 4A (a). Systematic analysis of metastasis showed that it was also positive to YY1 (72 ± 12% of positive sections), VEGF (64 ± 10% of positive sections), and CXCR4 (81 ± 15% of positive sections) immunostaining (Fig. 4A,, b). From a qualitative point of view, reduction of SaOs-2 metastasis induced by pretreatment with T22 (Supplementary Table S2; Fig. 4C,, a) promotes an enhanced positive immunostaining to YY1 and a negative immunostaining to the CXCR4 antibody (Fig. 4C,, b). Four mice from a group of 10 injected with clone 9 cells developed a diffuse micrometastasis with a mean area 10-fold smaller than that of control SaOs-2 mice (Supplementary Table S2; Fig. 4B,, a compared with A, a) positive to CXCR4 (Fig. 4B,, b) but negative to YY1 antibody staining (Fig. 4B , b). The treatment with T22 peptide was statistically ineffective on these mice (Supplementary Table S2).

To obtain detailed information on intrametastatic vascularization, we serially counted the new vessel formation within metastasis. As shown in Fig. 5A, the metastases obtained with clone 9 were less positive to CD34 and CD31 immunostaining than those with SaOs-2 (P < 0.001). The treatment with T22 peptide reduced the positive immunostaining by only 10% in metastases from mice receiving clone 9 (P < 0.01), compared with those from mice receiving SaOs-2 wild-type cells (P < 0.01; Fig. 5A). Metastases obtained with infusion of the clone 9 cells were 40% less positive to the CD31 antibody than those induced by SaOs-2 wild-type cells (P < 0.001; Fig. 5B). The treatment with the T22 peptide reduced by 50% the number of intrametastatic positive serial sections to CD31 (P < 0.001) in mice receiving clone 9 metastases and by 20% (P < 0.01) in those receiving SaOs-2 cells (Fig. 5B). In metastatic tissues, specific markers of metastasis (MTA1 and MTA2) were not significantly changed in our experimental conditions (Fig. 5C). To verify whether YY1 regulated directly or indirectly the expression of the master proangiogenic factor VEGF and the receptor VEGFR3, which are implied in lymphatic and microvessel formation (26), we analyzed a protein extract from metastases with VEGF antibody that was able to recognize the subforms of VEGF. As shown in Fig. 5D, only the VEGF-D subform was reduced in clone 9 + T22 metastasis and in mice receiving SaOs-2 + T22. The immunohistochemistry indicated that 41 ± 8% of clone 9 + T22 metastasis (P < 0.001) and 35 ± 6% of SaOs-2 + T22 metastasis (P < 0.003) were positive to VEGF-D. In contrast, in our experimental conditions, VEGFR3 expression was not altered significantly by different treatments. We were not able to appreciate differences among other subforms of VEGF in silenced YY1 metastasis and control, whereas in silenced cell lines we observed a reduced expression of VEGFA compared with SaOs-2 as shown in Fig. 5D.

Genetic profile of siRNAYY1 and wild-type SaOs-2 cells. To start a broad examination of gene expression profile involved in YY1-mediated invasiveness in SaOs-2 cells, we performed repeated microarray experiments on the siRNAYY1 pool of clones compared with the SaOs-2 (Table 1). This approach was not followed in orthotopic murine model due to a variety of cells (human and mouse cells; selective laser capturing of cells was not feasible in small micrometastasis of mouse lungs; refs. 23, 24). Our data show that 401 genes are differentially regulated in a significant way in siRNAYY1 clones compared with SaOs-2 wild-type cell control: 326 genes were up-regulated significantly, whereas the remaining 85 genes were down-regulated (Table 1). The up-modulated genes were grouped according to cellular component and biological functions. Among the up-regulated and down-regulated genes, we focused our attention on extracellular matrix components (collagen, fibronectin, and laminin) and on those involved in biological functions, such as proliferation, cytoskeleton reorganization, differentiation, and angiogenesis. Some random genes were analyzed by semiquantitative RT-PCR to confirm the microarray results (shown in Supplementary Fig. S1). These data were consistent with the genomic profile.

We show that YY1 silencing reduces cell growth in adherent, semisolid medium conditions and adhesion to different substrates, particularly to collagen (27). Moreover, we were not able to select clones with complete ablation of YY1 protein, in agreement with other reports, suggesting that it is necessary for cell survival (28). Our array data, as well as a previous study in mouse fibroblasts (28), indicate a loss of several genes involved in cell cycle and adhesion-migration pathways in silenced YY1 clones, suggesting that YY1 has a role in cell survival and in the acquisition of aggressive phenotype of the cells. Neoplastic cell migration is driven by CXCR4/SDF-1 axis in osteosarcoma cells (14–16). The dosage of CXCR4 expression in these cells indicated that the mRNAs were lower in silenced clones than in control (SaOs-2 cells); however, the protein was a little higher, indicating that YY1 is positively correlated with CXCR4 mRNA, whereas increased cell surface expression of CXCR4 may also be the result of altered regulation independent of transcription/translation mechanism. Ubiquitination of CXCR4 is a modification regulating the expression of CXCR4 posttranslationally (11).

Here, we show that silencing of YY1 reduces (4-fold) the cell migration mediated by SDF-1 that was specifically inhibited by the addition of T22 blocking peptide of CXCR4, reducing the transcription and activity of MMPs. Indeed, metalloproteinase assay revealed that the MMP2 mRNA was very low in clones silenced for YY1 compared with SaOs-2 wild-type cells. MMPs are a family of Zn²⁺-dependent endopeptidases, which play important roles in tumor angiogenesis and can be activated via CXCR4/SDF-1 pathway (25, 29). In particular, MMP-2 seems to be activated during angiogenesis by αvβ3 integrin, degrading type IV collagen (a major component of the basement membrane) and favoring tumor progression (30). In contrast, CXCR4 expression was higher in siRNAYY1 clones than in SaOs-2. All together, these data suggest that SDF-1 ligand promotes the migration of SaOs-2 cells but not of siRNAYY1 clones, which have a lower capacity to degrade Matrigel. However, in vivo, MMP activity synergizes with other growth factors, integrins, and the microenvironment. Osteosarcoma metastasizes to the lung, which is one of the preferential sites for SDF-1 production. Therefore, a specific regulation of MMPs or CXCR4 can occur at this site.

In addition, we have investigated whether the reduction of YY1 protein in osteosarcoma cells can interfere with their metastatic implantation, angiogenesis, and CXCR4/VEGF pathways. Using angioreactors, we directly quantified the number of newly formed vessels that the implantation of siRNAYY1 clone 9 cells was able to promote. We have shown that the silencing of YY1 interferes with formation of new vessels. Mice inoculated with silenced YY1 cells produced fewer vessels with lower lumen than SaOs-2 cell–bearing mice. These experiments indicate that the implantation of silenced YY1 cells was less effective for activating the tumor proangiogenic microenvironment than SaOs-2 cells. The fact that CXCR4 blocking peptide was ineffective in reducing the new vessel formation in vivo suggested that YY1 interfered with CXCR4/angiogenesis pathway downstream the receptor activation. Examination of lung metastases revealed that the silencing of YY1 reduced the metastatic implantation and tumor growth indeed (i.e., 9 of 10 SaOs-2–bearing mice developed metastases, whereas only 4 of 10 mice bearing silenced YY1 cells did). Consistently, the mean area of SaOs-2 metastases was 10-fold bigger than that observed in silenced YY1 mice and both were positive to CXCR4/VEGF pathways. The administration of the T22 peptide in SaOs-2–bearing mice significantly reduced (10-fold) the size of metastasis and slightly reduced (10%) the intrametastatic vascularization and VEGF-D expression, but it was not able to reduce their implantation significantly. In mice bearing silenced YY1 cells, the T22 peptide was ineffective on metastatic size, and it slightly reduced (10%) the VEGF-D protein expression and the intrametastatic vascularization. VEGF-D stimulates the growth of vascular lymphatic endothelial cells via the VEGFR3 signaling. VEGF-D is the most abundant isoform expressed in the lung and skin (17, 26, 31, 32), and it has been proposed that it has a role in tumor angiogenesis in lung (17, 26) and breast (26, 33). Moreover, a tandem action of CXCR4/VEGF is possibly present in prostate cancer (34) and renal carcinoma (35). We were not able to appreciate differences among VEGF subforms in vivo. However, in clone 9 cells, we observed a reduction in VEGFA compared with SaOs-2 cells. Moreover, we have observed directly in vivo that the YY1 silencing reduced the new vessels that are interfering with CXCR4/SDF-1 pathway.

Microarray analysis shows ∼400 genes differentially expressed during silencing of YY1 in SaOs-2 cells. Several of these genes are involved in cell cycle control and angiogenesis. Studies of proteomics are beyond the scope of this study; they are, however, performed presently in our laboratories to further investigate the meaning and causal relationship of these genes and the effects of their transcription on development of metastasis of osteosarcoma cells. Collectively, these data indicate that YY1 per se is involved in the positive regulation of CXCR4 transcription and is also a positive regulator of angiogenesis and migration pathway of osteosarcoma cells acting downstream of the CXCR4/SDF-1 pathway, suggesting that it can directly regulate directly CXCR4 but it is also an effector of CXCR4 signal transduction. Overall, our findings may have important therapeutic implications in terms of vascular targets in metastatic dissemination (8, 9, 36–38). However, we need to keep in mind that a complex alteration of cell cycle machinery also occurs during growth of malignant osteosarcoma (3, 6, 39, 40). Thus, molecular therapy should simultaneously target simultaneously angiogenesis on which depends cancer growth and major concurrent alterations of cell cycle of bone cancer cells depend.

Grant support: Italian Ministry of University and Research P.R.I.N./MIUR 2006 (PRIN 2006062153_002, “Meccanismi fisiopatologici di danno vascolare/trombotico e neoangiogenesi”; C. Napoli), Ricerca di Ateneo 2005-2006 from the II University of Naples (C. Napoli), Fondation Jerome Leyeune (Paris, France; C. Napoli), and Regione Campania (legge 5; A. Lanza and C. Napoli).

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.

We thank Prof. Francesco Bresciani for a critical reading of the manuscript, Prof. Ciro Abbondanza for helpful comments, Dr. Stefania Scala for recombinant SDF-1, and Dr. Chiara Botti for collaboration in initial pilot experiments of YY1 silencing in SaOs-2 cells.

This article is dedicated to the memory of Prof. Gaetano Salvatore, on the occasion of his tenth memorial anniversary.