Frequent Attenuation of the WWOX Tumor Suppressor in Osteosarcoma Is Associated with Increased Tumorigenicity and Aberrant RUNX2 Expression

The WW domain-containing oxidoreductase (WWOX) gene encodes a 46-kDa tumor suppressor that is altered in most human cancers (1, 2). WWOX spans the second most active common fragile site, FRA16D, frequently involved in cancer (3, 4). Loss of heterozygosity (LOH), homozygous deletions (HD), hypermethylation, and chromosomal translocations affecting WWOX have been reported in many malignancies. WWOX inactivation has been associated with more aggressive tumors and with poor prognosis (2, 5, 6). Recently, we have generated a targeted ablation of the Wwox gene in mice (7). Analysis of Wwox-mutant mice shows a bona fide tumor suppressor function of WWOX. More than 30% of Wwox-deficient juvenile mice develop periosteal osteosarcomas (OS; ref. 7). Importantly, molecular characterization of these tumors identifies RUNX2 as a partner of WWOX (8).

RUNX2 is an essential transcription factor for bone formation, and accumulative data have shown that its levels are upregulated in many malignancies, including OS and metastatic breast and prostate cancers (9–14). Our recent findings show that WWOX physically associates with RUNX2 and functionally suppresses its transactivating function in osteoblasts and in neoplastic cells (8). Furthermore, analysis of Runx2 mRNA expression in femurs of Wwox-deficient mice revealed its significant upregulation (8), suggesting that WWOX may regulate RUNX2 levels. Interestingly, osteoblasts isolated from Wwox-deficient mice exhibit an autonomous defect in differentiation (8).

OS is the most common primary malignant bone tumor in childhood and adolescence (15). This highly aggressive tumor usually involves long bones and frequently metastasizes to the lungs. Little is known of the etiology of OS and lesser still of the various interactions that occur between host and tumor cells to regulate the growth and progression of OS in vivo. OS is genetically heterogeneous, marked by multiple random chromosomal aberrations with only a few recurring deletions and amplifications (16, 17). Recently, it has been shown that mice with osteoblast-restricted loss of p53 and Rb uniformly develop OS (18). Molecular analyses of OS have also revealed alterations in transforming growth factor-β/bone morphogenetic protein, Wnt/β-catenin, and cell cycle–related pathways (10, 19–21). RUNX2 levels are increased in human OS (10, 12–14); however, the cause of this elevation is not known. Molecular markers predictive of OS tumor progression, metastasis, and response to chemotherapy are lacking, preventing development of more biologically tailored therapies for high-risk patients.

Here, we report immunohistochemical studies of WWOX and RUNX2 expression in a series of human OS samples recovered as biopsies from untreated primary tumors, bulk resections following chemotherapy, and excised metastases. We also describe the effect of overexpressing WWOX in human and in murine OS cell lines on the proliferation, migration, and invasion in vitro of the cell lines and on their xenograft growth in nude mice. Our findings suggest that reduction of WWOX expression is a common event in human OS and that its expression could serve as a prognostic and therapeutic indicator for assessing OS progression.

OS-derived cell lines were obtained from the American Type Culture Collection; human MG-63 cell line was a gift from Prof. Zvi Bar Shavit (Hebrew University), and MLOY4 cells were received from Prof. Linda Bonewald (University of Kansas). Cells were cultured in RPMI 1640 supplemented with 10% fetal bovine serum.

The WWOX adenovirus transduction has been previously described (22). Ad-GFP virus served as a control. Cells were transduced with Ad-WWOX or Ad-GFP at appropriate multiplicities of infection (MOI), and transduction efficiency was assessed by visualization of green fluorescent protein (GFP)–expressing cells. To generate a lentiviral WWOX vector, human WWOX cDNA was cloned in lentiviral vector containing neomycin as the selection marker (gift from Dr. Ittai Ben-Porath, Hebrew University) using the Gateway cloning system (Invitrogen).

Animal studies were done under an approved protocol according to institutional guidelines. KHOS, HOS, and K7M2 OS cells were infected with Ad-WWOX or Ad-GFP at a MOI of 100. Twenty-four hours after infection, 10⁷ cells were injected s.c. into the flanks of 6-week-old female nude mice (five mice per group; Charles River Laboratories). Five control mice were injected with 10⁷ uninfected cells. Animals were monitored daily, and tumor sizes were measured every 5 days. At the end point (day 28), animals were sacrificed, tumors were weighed, and tumor volumes were calculated as previously described (22).

Human OS tissue sections were obtained following Institutional Review Board approval. For each tumor, immunohistochemistry was performed on one representative formalin-fixed, paraffin-embedded section chosen by review of the H&E-stained slides. Polyclonal anti-WWOX antibody (1:5,000 dilution; ref. 23) or monoclonal anti-RUNX2 antibody (1:100 dilution; ref. 12) was used. Immunohistochemical staining and scoring of WWOX and RUNX2 staining were determined by at least two pathologists. For WWOX, tumors were grouped into three categories (strong, reduced, and absent) based on the intensity of cytoplasmic staining in all tumor cells present on the slides. The “strong” category was assigned when the intensity was equivalent to that found in osteoblasts and chondrocytes from healthy tissues. The “reduced” category was used for tumors with clearly diminished intensity, and the “absent” category for tumors with no WWOX immunostaining. When present, osteoblasts in reactive bone surrounding tumors and chondrocytes in bronchial cartilage served as internal positive controls. Nuclear RUNX2 staining was scored as either positive or negative. Tumors with only scattered RUNX2-positive cells (<2% of all tumor cells) were considered negative.

Results of in vitro and in vivo experiments were expressed as mean ± SD or SE. Fisher's exact test, Student's t test, and 95% confidence interval (CI) based on exact binomial CI were used to compare values of test and control samples. P < 0.05 indicated significant difference.

Detailed Materials and Methods are provided as Supplementary Data.

We previously reported that ∼30% of Wwox-deficient mice developed spontaneous OS (7). This incidence was based on sections through paraffin-embedded bone. Because only few serial slides contain tumor, we developed a screening approach using a more sensitive detection method by micro–computed tomography (μCT) imaging of intact limbs (Supplementary Fig. S1A). Irregular protrusions on the endosteal or periosteal sides of the cortex seen on μCT were then subjected to histologic assessment (Supplementary Fig. S1B). Using this approach, 100% of Wwox-deficient mice had developed OS by 18 days of age and 68% of mice had bilateral involvement (Supplementary Table S1). These new findings strengthen the hypothesis that WWOX deficiency can contribute to human OS.

To assess the clinical significance of WWOX protein expression in human OS, 83 OS samples from 51 patients (Supplementary Table S2) were analyzed by immunohistochemistry and compared with 12 normal bone and cartilage tissues (Table 1A). These included 34 pretreatment biopsies, 34 posttreatment resections, and 15 posttreatment metastasis resections. Representative examples of OS with normal, reduced, and absent WWOX immunoreactivity are shown in Fig. 1A. WWOX expression was uniformly strong in the cytoplasm of the 12 normal bone and cartilage specimens (Supplementary Fig. S2; Fig. 1A). In OS specimens, strong WWOX expression was detected in 42% (35 of 83), whereas 58% (48 of 83) exhibited absent or reduced (grouped as Altered) WWOX immunoreactivity (P < 0.0001; Table 1A). These data indicate that WWOX expression is significantly altered in OS tumors compared with normal bone.

After biopsy diagnosis, OS is treated with a chemotherapeutic induction regimen of cisplatin, doxorubicin, and high-dose methotrexate before surgical resection (24). To determine the effect of chemotherapy on WWOX expression, the immunohistochemical expression of WWOX was compared in pretreatment biopsies and posttreatment resection specimens from 16 OS patients. WWOX expression increased in 44% (7 of 16) of the paired samples, 5 of which were initially WWOX negative (95% CI, 19.75–70.12; Fig. 1B; Supplementary Table S3). Another 44% (7 of 16) of the paired samples showed no change in WWOX; however, all expressed WWOX in the pretreatment biopsies. WWOX decreased in postchemotherapy specimens in 22% (2 of 16) of the paired samples. These results suggest that chemotherapy, which induces tumor cell normalization, is accompanied by restoration of WWOX expression. Interestingly, all samples where WWOX increased had a poor response to chemotherapy (<90% necrosis), whereas 67% of the remaining paired samples had a favorable response (Supplementary Table S4).

Because OS frequently metastasizes, we determined whether WWOX expression in metastases was different than in primary OS by comparing immunohistochemical expression in paired samples from nine patients who had a pretreatment OS biopsy and a posttreatment metastasectomy. Nearly half (four of nine) of paired specimens showed reduced or absent WWOX expression in the metastasis compared with the primary tumor (95% CI, 13.70–78.80; Fig. 1C; Supplementary Table S5). WWOX levels were unchanged in three of the paired samples (two already had reduced WWOX levels in the primary tumor), whereas two metastases exhibited higher WWOX expression compared with the primary tumor. These results suggest that decreased WWOX levels are associated with a more aggressive tumor cell at the metastatic site.

To determine the expression of WWOX in OS cells, we examined mRNA levels and protein expression in five human and three mouse OS cell lines. Four human OS cell lines (MG-63, KHOS, SAOS2, and HOS) exhibited significantly lower WWOX expression than in hBM, WWOX-positive MCF-7 breast cells, and murine preosteoblastic MC3T3 cells (Fig. 2A). WWOX mRNA expression in these human OS cells was comparable with MDA-MB-231 cells, a breast cancer cell line that expresses very low levels of WWOX (25). By contrast, Wwox mRNA expression in the K7M2 mouse OS cell line was comparable with mouse bone marrow, MC3T3, and MLOY4 normal osteocytes (Fig. 2B). Protein expression was barely detected in the four human cell lines compared with MCF-7 cells, which express abundant WWOX (Fig. 2C). The K7M2 mouse OS cells exhibited comparable WWOX protein levels with the mouse preosteoblastic MC3T3 cells (Fig. 2C). These results indicate that WWOX expression is altered in most human OS cells.

Because four of five human OS cell lines exhibited low endogenous WWOX expression, we determined whether overexpressing WWOX would alter their growth. We overexpressed WWOX in three OS cell lines using Ad-WWOX (22) or Ad-GFP at a MOI of 100. Immunoblot analysis 72 hours after infection confirmed WWOX overexpression in all Ad-WWOX–infected cells (Fig. 3A). Next, we measured cell proliferation by the 2,3-bis[2-methoxy-4-nitro-5-sulfophenyl]-2H-tetrazolium-5-carboxanilide inner salt (XTT) assay. Two OS cell lines (KHOS and HOS) that have very low endogenous WWOX levels exhibited reduced proliferation following Ad-WWOX infection compared with Ad-GFP infection (Fig. 3B). In contrast, K7M2 cells, which endogenously express higher levels of WWOX, had no change in proliferation following WWOX overexpression (Fig. 3B).

Previous studies have shown that ectopic WWOX expression promotes apoptosis (22). We therefore used propidium iodide and flow cytometry to assess whether WWOX overexpression induces apoptosis in the OS cell lines. A sub-G₁ population was present in Ad-WWOX infected in KHOS and HOS cells but not in K7M2 cells (Supplementary Fig. S3). At 72 hours after Ad-WWOX infection (MOI, 100), but not after Ad-GFP infection, the sub-G₁ fraction was increased ∼6-fold in KHOS cells and ∼3-fold in HOS cells.

We next measured tumorigenicity of OS cells infected with Ad-WWOX or Ad-GFP by inoculating nude mice with 1 × 10⁷ cells. At 28 days after injection, three of five and two of five mice inoculated with Ad-WWOX–infected HOS cells and KHOS cells, respectively, displayed no tumors, whereas all Ad-GFP–infected mice had tumors (data not shown). The average volume of tumors at day 28 after inoculation with Ad-WWOX–infected cells showed significant suppression of tumor growth in KHOS and HOS cells, but not in WWOX-positive K7M2 cells, compared with tumors from Ad-GFP treatment (Fig. 3C).

Because OS cells often metastasize, we used lentiviral vectors to generate stable clones of HOS and KHOS cells expressing WWOX or empty vector (EV) and then measured their proliferation and cell migration characteristics. HOS-WWOX and KHOS-WWOX stable clones had increased WWOX expression compared with HOS-EV and KHOS-EV clones (Fig. 4A; Supplementary Fig. S5A). We performed XTT, colony formation, and soft agar assays and found that HOS-WWOX and KHOS-WWOX cells had reduced proliferation, survival, and anchorage-independent cell growth compared with HOS-EV and KHOS-EV clones and their respective parental cell lines (Fig. 4B; Supplementary Figs. S4 and S5).

HOS-WWOX cells had low cell motility compared with HOS-EV and HOS cells in a wound-healing assay (Fig. 4C). In a Matrigel invasion assay, HOS-EV clones behaved similar to HOS cells, whereas HOS-WWOX cells exhibited reduced invasiveness (Fig. 4D). Similar results were observed when KHOS-WWOX stable clones were compared with KHOS-EV clones (Supplementary Fig. S5).

We reported recently that WWOX suppresses the transactivation function of RUNX2 (8); therefore, we set out to determine whether WWOX tumor suppressor function in OS is associated with elevated levels of RUNX2. First, we examined RUNX2 in femurs of Wwox-deficient mice and found elevated immunohistochemical levels of RUNX2 compared with wild-type littermate control mice (Supplementary Fig. S6). Next, by Western blot analysis in cell lines, we also found an inverse correlation between levels of WWOX and RUNX2 in several lines (Fig. 5A). The breast cancer cell line MCF-7 has high levels of WWOX and absent RUNX2, whereas WWOX levels are low and RUNX2 is expressed in the highly metastatic cell line MDA-MB-231. Similarly, murine cell lines MC3T3 and K7M2 express WWOX and have reduced RUNX2 expression, whereas human OS cell lines HOS, SAOS2, and KHOS have reduced WWOX expression and elevated RUNX2 expression. Interestingly, the inverse correlation between WWOX and RUNX2 was not observed in two less differentiated OS cell lines, MG-63 and U2OS. Importantly, overexpression of WWOX in the lentivirus-transduced HOS-WWOX cells caused a significant reduction in RUNX2 mRNA and protein as assessed by quantitative reverse transcription-PCR (qRT-PCR) and Western blot (Fig. 5B and C). Similarly, downstream target genes of RUNX2, matrix metalloproteinase-9 (MMP9), type I collagen (COL1A1), and, to a lesser extent, osteocalcin (OC), showed marked downregulation in HOS-WWOX cells compared with HOS-EV cells (Fig. 5B). These data suggest that WWOX expression may contribute to OS formation, at least in part, through regulating RUNX2 expression.

Because the majority of human OS examined express elevated levels of RUNX2 (10, 14), and our findings showed that WWOX expression is attenuated in OS, we tested whether WWOX and RUNX2 expression are inversely correlated in human OS specimens. Immunohistochemical analyses of 56 of our OS cases (Supplementary Table S2) revealed that 60% (12 of 20) of pretreatment biopsies were positive for RUNX2 (Table 1B). Only 16% (4 of 25) of posttreatment resections were positive for RUNX2, indicating that the majority of resected samples had lost this highly metastatic marker. Paired pretreatment biopsy and posttreatment resections were available for 12 OS patients. Eight of the biopsies were RUNX2 positive and all (100%) became RUNX2 negative after treatment, whereas one in four of the biopsies that were RUNX2 negative became RUNX2 positive after treatment (Supplementary Tables S4 and S6), suggesting a poor outcome. A similar comparison was performed using paired pretreatment biopsies and resected metastases from seven OS patients. Interestingly, among the seven patients for whom pretreatment biopsies and metastasectomy specimens were available, all pretreatment biopsies were RUNX2 positive and five remained RUNX2 positive in the metastases (Supplementary Table S7), consistent with poor prognosis.

Although WWOX and RUNX2 results suggested an inverse association in human OS, this was not evident when paired comparisons were performed on the 56 available cases (Supplementary Table S8). In pretreatment biopsies, 78% (seven of nine) of WWOX strong tumors were also RUNX2 positive. In WWOX reduced or absent biopsies, 55% (6 of 11) were also RUNX2 negative. Evaluation of both WWOX and RUNX2 levels following chemotherapy revealed that, of the 6 of 12 patients with increased WWOX following chemotherapy, 3 (50%) had a concomitant decrease in RUNX2. Two others were RUNX2 negative in both biopsy and resection such that 83% (five of six) of posttreatment specimens with increased WWOX were RUNX2 negative (Supplementary Table S4). Two of the three metastases with decreased WWOX also had decreased RUNX2 despite findings that most of the metastases are RUNX2 positive (Supplementary Table S9).

We find that WWOX expression is absent or reduced in 58% of OS patient specimens (P < 0.0001) and in the majority of human OS cell lines. In some OS patients, chemotherapy restored WWOX levels and the metastases had reduced levels. The observed alterations in WWOX expression were accompanied, at least in part, by aberrant RUNX2 expression, a factor known to be expressed in OS (10, 14). Our data also revealed that WWOX expression suppresses tumorigenicity both in vitro and in vivo in nude mice and that ectopic WWOX expression diminishes tumor progression of metastatic OS cell lines. These findings provide the first evidence that WWOX expression is clinically and therapeutically significant in OS.

Consistent with other types of human cancer (1, 2, 5), we observed that a significant proportion of OS samples exhibited either markedly reduced (41%) or absent (17%) WWOX expression (Fig. 1; Table 1A). WWOX protein and mRNA levels are significantly reduced in many human OS cell lines compared with control cells (Fig. 2). Previous studies have shown that hypermethylation of the regulatory region of WWOX is associated with low or absent protein expression (5); however, we could not confirm this in OS cell lines (data not shown). Alteration of WWOX in cancer has also been shown to be associated with LOH, HD, and enhanced protein degradation (5). Tumor suppressor genes at common fragile sites are frequently inactivated early in neoplastic progression (26). It is thus possible that inactivation of WWOX could occur during the extensive proliferation necessary for early bone growth, contributing to development of OS in childhood.

Because loss of WWOX results in OS formation in mice (ref. 7; Supplementary Fig. S1), and reduced levels are present in human OS cell lines and tissues, we determined whether ectopic expression of WWOX could reverse malignancy. In restoring WWOX expression in two WWOX-reduced OS cell lines by infection with Ad-WWOX, we observed a decreased tumorigenicity in vitro, as indicated by reduced proliferation (Fig. 3B) and increased apoptosis (Supplementary Fig. S3), and in vivo, as indicated by marked reduction of tumor growth in immunocompromised mice (Fig. 3C). Thus, ectopic expression of WWOX in OS cells can potentially reverse malignant properties despite other cancer-associated genetic alterations that have accumulated in these cell lines. Decreased tumorigenicity was not observed in K7M2 OS cells that express WWOX, suggesting that WWOX overexpression may not affect the cellular phenotype in WWOX-sufficient cells.

We previously showed that WWOX, via its first WW domain, associates with RUNX2 and suppresses its transactivation of the OC promoter (8). Increased RUNX2 levels have been reported in human OS (10, 12–14). Our data show that RUNX2 protein and mRNA levels were also high in femurs of Wwox-deficient mice and in some OS cell lines with low endogenous levels of WWOX, suggesting an inverse correlation with WWOX (Supplementary Fig. S6; ref. 8). RUNX2 autoregulatory mechanisms have been described in osteoblasts (27, 28), and several lines of evidence indicate that RUNX2 expression is regulated by different tumor suppressors and oncoproteins implicated in the pathogenesis of OS, including pRb, p53, and MDM2 (18, 29, 30). Additionally, WW domain-containing proteins such as YAP (31) and the potent activator TAZ (32) interact with the same protein domain as WWOX. Therefore, it is possible that when WWOX is absent, reduced, or nonfunctional, other WW domain-containing proteins may regulate RUNX2 transcription, contributing to its increased levels in OS.

In human OS samples, we found that 60% of prechemotherapy OS biopsies were RUNX2 positive (Table 1B). Interestingly, when paired comparisons were performed on 56 human OS samples, an inverse relationship between WWOX and RUNX2 was not evident (Supplementary Table S8). More than half of WWOX altered primary tumors were also RUNX2 negative, and ∼80% that were WWOX strong were also RUNX2 positive. The strong inverse relationship found in vitro may reflect ex vivo selection of WWOX altered cell lines due to their enhanced growth properties, although we did observe an inverse correlation in vivo in Wwox-deficient mice. It is possible that the immunohistochemistry performed on fixed human tissue samples lacked the sensitivity to detect changes in WWOX and RUNX2 expression seen by qRT-PCR assays and Western blots, performed on cell lines. Alternatively, the relationship between WWOX and RUNX2 may be more complex in vivo, in part due to multiple additional aberrations that occur in the tumors. Support for this latter hypothesis is found in the less differentiated cell lines, which showed both WWOX and RUNX2 expression to be either reduced (MG-63) or preserved (U2OS; Fig. 5A). An expanded study will be required to explore this aberrant relationship in human OS.

After tumor staging, the most important predictor of disease-free survival in OS is responsiveness to chemotherapy. Ninety percent or greater tumor necrosis at resection (good response) is associated with a more favorable prognosis. Currently, there are no molecular markers to further stratify the poor responding patients for additional therapy (33–35). We examined WWOX immunoreactivity in paired biopsy and resection specimens and found that in nearly half of postchemotherapy resections, WWOX expression was increased (Supplementary Table S3). None of the resections having increased WWOX expression had high (>90%) rates of tumor necrosis. At present, our sample size is too small to assess the significance of this observation. However, we speculate that WWOX expression might serve as another predictor of tumors with favorable and less favorable prognoses (Supplementary Table S4).

Metastasis, a hallmark of malignancy (36), is found in up to 20% of patients with OS at initial diagnosis and is associated with a poor prognosis (33, 35). Our analysis revealed that nearly half of metastases had reduced or absent levels of WWOX compared with their primary tumor (Supplementary Table S5). We also showed that stable restoration of WWOX in metastatic HOS and KHOS cell lines inhibited anchorage-independent growth and colony formation (Fig. 4), and reduced migration and Matrigel invasion in vitro (Fig. 4; Supplementary Fig. S5), suggesting that loss of WWOX expression might be associated with development of metastases.

Intriguingly, the majority of OS cases experienced a loss of RUNX2 following chemotherapy (Supplementary Tables S4 and S6). Nearly all cases where WWOX increased following chemotherapy were also RUNX2 negative, suggesting a normalization of the WWOX-RUNX2 relationship with chemotherapy. Another significant outcome of our study is that the majority of metastatic tumors expressed RUNX2 (Supplementary Tables S7 and S8), consistent with its oncogenic potential (9). Similar to the primary tumors, an inverse relationship was not found between WWOX and RUNX2 in most resections and metastases. It is possible that failure to restore the inverse relationship between WWOX and RUNX2 levels with chemotherapy identifies cases with an inherent predisposition toward metastasis. Larger case series will be necessary to evaluate whether assessing WWOX and RUNX2 expression will be a useful prognostic tools in OS.

In conclusion, WWOX expression is lost or reduced in a large portion of human OS and seems accompanied by an aberrant relationship with RUNX2. Ectopic expression of WWOX in human OS cells suppressed cancer cell growth, diminished migration and invasion in vitro, and suppressed xenograft tumor growth. Tumors that increased WWOX levels and lost RUNX2 expression following chemotherapy had poor necrosis responses, suggesting that a subset of poor responders may have improved prognosis due to tumor cell normalization. This normalization was not found in metastases, which largely retained patterns seen in the primary tumors despite chemotherapy. Our findings suggest potential prognostic value in examining WWOX and RUNX2 in OS patients.

No potential conflicts of interest were disclosed.

We thank Dr. Ittai Ben-Porath for providing the lentiviral vector and Prof. Kay Huebner and Dr. Muller Fabbri (Ohio State University) for polyclonal anti-Wwox antisera and Ad-WWOX virus.

Grant Support: American-Israeli Bi-national Science Foundation (R.I. Aqeilan and J.L. Stein), Israeli Cancer Research Funds (R.I. Aqeilan), Israeli Cancer Association (R.I. Aqeilan), Alex's Lemonade Foundation (R.I. Aqeilan), Howard Hughes Medical Institute (M. Warman), and NIH grants P01 CA082834 and P01 AR048818 (G.S. Stein and J.B. Lian).

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.