Osteosarcoma is the most common malignancy of bone, and occurs most frequently in children and adolescents. Currently, the most reliable technique for determining a patients’ prognosis is measurement of histopathologic tumor necrosis following pre-operative neo-adjuvant chemotherapy. Unfavourable prognosis is indicated by less than 90% estimated necrosis of the tumor. Neither genetic testing nor molecular biomarkers for diagnosis and prognosis have been described for osteosarcomas. We used the novel nanoString mRNA digital expression analysis system to analyse gene expression in 32 patients with sporadic paediatric osteosarcoma. This system used specific molecular barcodes to quantify expression of a set of 17 genes associated with osteosarcoma tumorigenesis. Five genes, from this panel, which encoded the bone differentiation regulator RUNX2, the cell cycle regulator CDC5L, the TP53 transcriptional inactivator MDM2, the DNA helicase RECQL4, and the cyclin-dependent kinase gene CDK4, were differentially expressed in tumors that responded poorly to neo-adjuvant chemotherapy. Analysis of the signalling relationships of these genes, as well as other expression markers of osteosarcoma, indicated that gene networks linked to RB1, TP53, PI3K, PTEN/Akt, myc and RECQL4 are associated with osteosarcoma. The discovery of these networks provides a basis for further experimental studies of role of the five genes (RUNX2, CDC5L, MDM2, RECQL4, and CDK4) in differential response to chemotherapy.
Citation: Martin JW, Chilton-MacNeill S, Koti M, van Wijnen AJ, Squire JA, Zielenska M (2014) Digital Expression Profiling Identifies RUNX2, CDC5L, MDM2, RECQL4, and CDK4 as Potential Predictive Biomarkers for Neo-Adjuvant Chemotherapy Response in Paediatric Osteosarcoma. PLoS ONE 9(5): e95843. doi:10.1371/journal.pone.0095843
Editor: David Loeb, Johns Hopkins University, United States of America
Received: September 26, 2013; Accepted: March 31, 2014; Published: May 16, 2014
Copyright: © 2014 Martin et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was funded by the Canadian Cancer Society through grant CCRI-020247. Additional support was provided by National Institutes of Health grant AR049069 (to AvW). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Osteosarcoma is the most common primary malignant bone tumor arising from bone in children and adolescents. The tumor is very rare with an incidence approaching 5 per million per year . Tumors arise from mesenchymal cells predominantly in the metaphyses of the distal femur, proximal tibia, and proximal humerus adjacent to epiphyseal growth plates . In rare cases, osteosarcomas affect the axial skeleton and other non-long bones . Chemotherapy followed by surgical resection is the standard treatment for high-grade osteosarcoma, and the current drug regimen is a combination of high-dose methotrexate, doxorubicin, cisplatin, and ifosfamide.
Histopathologic examination to estimate tumor necrosis following neo-adjuvant pre-operative chemotherapy is currently one of the most reliable tools for response evaluation and prognostication. Unfavorable response, corresponding to a bad prognosis, is indicated by less than 90% estimated necrosis of the tumor following neo-adjuvant chemotherapy . Unlike other paediatric cancers, there are few consistent genomic translocations, amplifications, or deletions in osteosarcoma that are useful for clinical diagnosis/treatment . Similarly, a large number of gene products have potential for driving oncogenesis or disease progression in osteosarcoma . This complex biology of osteosarcoma has limited the identification of reliable molecular biomarkers for tumor classification or therapeutic targeting.
Definitive diagnosis of osteosarcoma requires the presence of immature osseous matrix around neoplastic cells, which are often osteoblast-like. The predominant epithelial mesenchymal lineages define the tumor subtype based on the main form of extracellular matrix observed: osteoblastic osteosarcoma (osteoid matrix), chondroblastic osteosarcoma (cartilaginous tissue), or fibroblastic osteosarcoma (fibrous tissue). The histological subtype may define specific molecular pathways involved in osteosarcoma development and progression . The high level of genetic and cytogenetic heterogeneity of this tumor, both between patients and within the tumors themselves , necessitates specific and personalised treatment approaches to improve outcomes.
Paediatric osteosarcoma represents a challenge to cancer treatment teams and research groups alike, a situation that is contrary to other sarcoma types for which expression signatures exist . In fact, a large proportion of other sarcomas are characterised by a single dominant-acting fusion protein encoded by a disease-specific chromosome translocation, while osteosarcoma cells possess cytogenetically complex karyotypes with no such consistent translocations . Scott et al. used a comparative biology approach to discover molecular subtypes of human osteosarcoma after studying profiles of canine osteosarcoma . RT-PCR- and gene expression microarray-based studies of paediatric osteosarcoma have previously been used to investigate disease-specific expression patterns and signatures –.
Our previous work revealed significant changes in a number of genes involved in tumor suppressive pathways, cell cycle control, and oncogenic mechanisms . In the present study, candidate genes were selected based on our previous work, as well as on the published reports on gene products with potential for involvement in osteosarcoma development. NanoString nCounter Technology, which has been used previously to classify other tumors , was used to determine expression levels of RNA from our cohort of 32 osteosarcoma patients. The nanoString Gene Expression Assay is a high-sensitivity, multiplexed method utilizing specific molecular bar codes for the detection of mRNAs that eliminates any enzymatic reactions . An analysis of the interaction of the most prominent biomarkers in this study with some of the other established oncogenic drivers in osteosarcoma was performed to determine which regulatory networks may underlie the varying responses to neo-adjuvant chemotherapy in this cohort.
The 32 patient cohort sample used in this study were obtained according to the guidelines and approval of the Sick Kids Hospital Research Ethics Board. Informed written consent to participate in this study was obtained from the patients, or in the case of young children, their next of kin, caretakers, or guardians on their behalf.
Paediatric Osteosarcoma and Normal Human Osteoblast Samples
Forty sporadic paediatric osteosarcoma tumor samples derived from a cohort of 32 patients were taken from pre-chemotherapeutic biopsies or from surgical resection specimens. All specimens were flash-frozen following the biopsy procedure or surgical resection. All surgical resection specimens had been exposed to the same standard regimen of neo-adjuvant chemotherapy, including methotrexate, doxorubicin, cisplatin, and ifosfamide. For eight patients, the initial diagnostic biopsy samples, which were naïve to neo-adjuvant chemotherapy, and a matched post-chemotherapy resection sample, collected on average 3.8 months after the first sample, were analysed. These eight pairs of RNA samples were used in a subset analysis to calculate the relationships of expression changes that occurred during the neo-adjuvant chemotherapy. Chemotherapy response of <90% necrosis was used to identify those tumors with a bad response following neo-adjuvant chemotherapy, and these values, along with other patient characteristics, are summarised in Table 1. Normal human osteoblasts isolated from surgical bone resections from five healthy individuals were obtained from PromoCell (Heidelberg, Germany) and combined and run as a single pooled sample. Total RNA was extracted from the tissues using the TRIzol Reagent method according to the manufacturer’s protocol (Invitrogen) and quantified using the Bioanalyzer (Agilent Technologies). Total RNA from normal human osteoblasts and osteosarcoma cell lines was retrieved as described previously . All aliquots were diluted to a final concentration of 20 ng/µL.
nanoString nCounter Assay
The nanoString nCounter gene expression system (nanoString Technologies) was used for expression profiling of the osteosarcomas and normal human osteoblasts. Details of the system are described elsewhere . Briefly, unique multiplexed probes were made with two sequence-specific probes per target mRNA. Two probes were constructed complementary to a 100-base target region. The capture probe comprised a target-specific oligonucleotide coupled to a short sequence linked to biotin. The reporter probe consisted of a second target-specific oligonucleotide linked to a unique chain of dye-labelled RNA segments for detection by the system. Our nCounter code set consisted of 21 probes, including 18 test probes derived from 17 distinct genes (RUNX2 comprised P1 and P2 transcripts) and three control genes (Table S1). Each sample was hybridised in duplicate or triplicate using 100 ng total RNA per reaction, in addition to the capture and reporter probes, as previously described .
Development of Candidate Gene List and nanoString Code Set Design
We selected 17 candidate genes for this study based on published reports describing gene products with the potential for involvement in osteosarcoma development, and based on our own findings (Table 2). The literature we considered, included gene copy number and gene expression microarray experiments, in addition to functional assays of genes, in models of osteosarcoma. In addition we performed pathway analysis using Ingenuity Pathway Analysis (IPA) to delineate overrepresented gene networks in the candidate genes associated with osteosarcoma oncogenesis. IPA employs the Fisher’s exact test to determine the relationship between the input dataset and canonical pathways with associated biofunctions (Ingenuity Pathway Analysis system; http://www.ingenuity.com/). Statistically significant overexpression in osteosarcoma tumors relative to normal osteoblasts has been detected previously for RECQL4, RUNX2, and SPP1 , , as is the case for amplification-related overexpression of CDC5L and RUNX2 osteosarcoma specimens . We included probes for each of the two RUNX2 transcript isoforms in the codeset. RUNX2_P1 captures expression of the normal osteoblast-specific version of RUNX2 , whereas RUNX2_P2 captures expression of RUNX2 during earlier stages of development. The latter version is also highly expressed in tumors, such as osteosarcomas . Unless otherwise specified as RUNX2 (P1), all of the reported expression of RUNX2 refers to the P2 transcript. Over-expression of the protein products of FOS, MYC, MDM2, CDK4, SPARC, and BCL2L1 have also been associated with osteosarcoma and have well-described tumorigenic potential –. On the other hand, a high frequency of genetic inactivation and copy number loss in osteosarcoma has been documented for TP53, CDKN2A, RB1, PTEN, and WWOX –. We included CDKN1A and CDKN1C in our analysis because of their roles in TP53 and RB1-mediated control of cellular proliferation . We selected HMBS, MT-ATP6, and MT-CO1 as housekeeping controls for our experiments because of validation in previous experiments –.
All data analysis was performed using the nSolver Analysis Software (nanoString Technologies). Briefly, counts are normalised for all target RNAs in all samples based on the positive control probes to account for differences in hybridisation efficiency and post-hybridisation processing, including purification and immobilisation of complexes. The software calculates the geometric mean of each of the controls for each sample to estimate the overall assay efficiency. Subsequently, mRNA content normalisation was performed using the housekeeping “calibration” genes, MT-ATP6, MT-CO1, and HMBS. Values <0 were blanketed and considered equal to 1 to facilitate downstream statistical analyses. All expression data can be accessed through GEO (Accession Number GSE45275). Following normalization (Table S2A and S2B) and removal of the samples with poor quality control data, the unpaired two-tailed Student’s t-test (Table S3) was applied to derive genes that had significant (p<0.05) differential expression in the two groups (25 with poor prognosis and 15 with good prognosis).
Correlations of Expression Change with Tumor Response to Chemotherapy
An unsupervised clustering was performed using Cluster 3.0 and Java tree view to determine the aggregation of the 40 RNA samples (in duplicates or triplicates) from the cohort in comparison to normal human osteoblasts and three osteosarcoma cell lines (Figure S1). An unsupervised analysis was repeated using only tumor samples from the cohort. The cluster map (Figure 1) shows the differential expression of the 17-gene probe code set from our nanoString panel, comparing the good to the poor responders. For the eight paired patient samples, only the initial pre-chemotherapy biopsies were used to examine expression levels of the gene set. Tumors with <90% necrosis in response to neo-adjuvant chemotherapy possessed significantly higher expression levels of: RUNX2, CDC5L, CDK4, and RECQL4; and significantly reduced levels of MDM2 (all p≤0.05) (Figure 2).
Patients with >90% tumor necrosis in response to chemotherapy (good response) are shown in blue and those with <90% (poor response) are shown in red. Sample numbers 1–36 were analyzed in triplicates whereas samples 41–116 were analysed in duplicates. Detailed data manipulation for the cohort is presented in Table S2B.
Unpaired Student’s t-test was applied to derive the genes that were differentially expressed in the two groups.
Gene Expression Changes Relative to Normal Human Osteoblasts
Using the duplicate measurements for each experiment, expression changes for 32 patient samples were normalised, averaged, and then ratios relative to normal human osteoblasts were calculated. Statistically significant up-regulated expression in tumors relative to human osteoblasts was detected for CDKN1C, FOS, MYC, RECQL4, RUNX2, SPARC, SPP1, and WWOX. Down-regulation was observed for CDKN1A and TP53 (Figure S2). Mean expression change and direction relative to normal osteoblasts match our previously published mRNA qRT-PCR analyses for CDKN1A, MYC, RUNX2, and SPP1 .
Gene Expression Levels Following Exposure to Neo-adjuvant Chemotherapy
For the eight patients with biopsies naïve to chemotherapy and matched resections, it was possible to perform a subset analysis to determine expression changes in resected samples relative to biopsies. mRNA expression was elevated for BCL2L1 (p<0.05), CDKN1A (p<0.05), MDM2 (p<0.05), PTEN (p<0.05), and WWOX (p<0.05). No significant differences were seen in the expression levels of RUNX2, CDC5L, CDK4, or RECQL4, between biopsy and resection specimens. Of the five genes for which differential expression in tumors was associated with a poor response to chemotherapy, only MDM2 levels changed after exposure to neo-adjuvant chemotherapy in this subset of eight patients.
To investigate the signalling relationship between the 17 selected genes in this study and the 31 candidate osteosarcoma driver genes from a previously published data set , standard pathway analysis using IPA software, was performed. Molecular interaction networks explored using IPA tools, allowing a maximum threshold of 35 nodes per network, revealed a total of 15 networks. The top two significant networks are shown in Figures 3A and B. Network 1 included 16 differentially regulated genes (score = 31), with signalling in RB1, TP53, PI3K, PTEN/Akt, MYC, and RECQL4 as the major over-represented gene networks. The interactions between the five-gene signature within signalling network 1 was strong, indicating that there were likely functional interactions of some of the genes with the core cellular regulation of cell cycle control. Network 2 included six genes (score = 8), which are associated with FOS, FAS, NFkB, and ERK1 signalling pathways. These functions are associated with extracellular signalling in the context of bone morphogenesis.
Panel A depicts the major over-represented network which has molecular relationships between some of the genes in code set and RB1, TP53, PI3K, PTEN/Akt, MYC, and RECQL4 interactions. Panel B depicts the second ranked network that shows interactions with FOS, FAS, NFkB, and ERK1 signalling pathways.
Paediatric osteosarcoma is a rare and complex tumor that has been studied extensively with respect to the genetic alterations that can be present . However, there is still no consistently reliable marker for clinicians to use in prognostication aside from the degree of tumor necrosis in response to chemotherapy, and no targeted therapies exist . Our objective in this study was to characterise a cohort of osteosarcoma tumors using nanoString technology. The nanoString nCounter is a digital expression analysis tool that is increasingly being used to detect and validate molecular signatures distinguishing subgroups of cancers . The efficacy of nanoString molecular bar code analyses relative to traditionally used technologies has been demonstrated by others .
Our nanoString expression analyses closely replicated our previous qRT-PCR analyses of an independent cohort of osteosarcoma specimens. Relative to normal human osteoblasts, we detected up-regulation of CDKN1C, FOS, MYC, RECQL4, RUNX2, SPARC, SPP1, and WWOX, with similar magnitudes of changes for MYC and RUNX2. The initial IPA analysis of our data, combined with that of Kuijjer et al. , demonstrates the potential for disruption of the MAPK/ERK signal transduction pathway via up-regulation of FOS, AP-1, and SPARC (Figure 3B). Relative to normal osteoblasts, changes in the network by aberrant expression of these genes would not only interrupt signal transduction through MAPK/ERK, but affect transcription, DNA stability, and apoptosis regulation.
Up-regulation of RUNX2, CDC5L, CDK4, RECQL4 and down-regulation of MDM2 was detected in the tumors that responded poorly to chemotherapy. There was also a statistically significant increase in RECQL4 expression in tumors relative to normal osteoblasts (Figure S2), a finding that corroborates previous studies from our research group . RECQL4 overexpression is associated with elevated genome instability in osteosarcoma , and genomic instability may, in turn, contribute to chemoresistance and poor prognosis . This gene encodes a DNA helicase important in DNA replication regulation during G(1) and S phases . Germ-line mutations in RECQL4 are associated with Rothmund-Thomson syndrome, which is linked to the development of several malignancies, including paediatric osteosarcoma . Somatic mutations of RECQL4 have not been detected in sporadic osteosarcomas, but the present study raises the possibility that genetic amplification may be the method by which RECQL4 dysfunction presents in sporadic tumors.
In contrast to RECQL4, there was minimal change in CDC5L or CDK4 expression between tumors and normal human osteoblasts (Figure 2), as was the case in a previous study by our group . Overexpression of CDC5L has been detected in osteosarcoma patient samples and osteosarcoma cell lines alike, and is a probable candidate oncogene within the 6p12-p21 amplicon commonly found in osteosarcomas . CDC5L is an essential component of the spliceosome complex, and elevated CDC5L shortens the G(2)-M cell cycle transition . Furthermore, CDC5L is important in the DNA damage response following exposure to genotoxic agents. It interacts with the checkpoint kinase ATR and is required for activating S-phase checkpoint effectors . The CDK4 protein regulates and promotes cell cycle progression through G1, and elevated levels are common in cancer . CDK4 amplification and overexpression are commonly found in low-grade  and dedifferentiated forms of osteosarcoma . CDK4 amplification and overexpression tend to be associated with better prognosis in the rarely occurring low-grade osteosarcomas  but is associated with poor prognosis in the majority of osteosarcoma cases. . Expression of MDM2 is known to inhibit TP53 transcriptional activation . Thus, differential expression of CDC5L, MDM2, and CDK4 could contribute to variation in cellular proliferation during osteosarcoma pathogenesis, especially when there is reduced expression of the tumor suppressors RB1 and TP53, or the TP53 target CDKN1A.
These results are also in keeping with a recent report from Kelly et al. , who used a similar general experimental approach to identify differentially expressed microRNA associated with chemotherapy response in osteosarcoma. Their study identified a small subset of miRNA and mRNA targets that impacted on some of the same genes and molecular pathways that we identified in this study. For example, there was a strong relationship between some of the gene targets identified in their study and bone morphogenesis proteins such as RUNX2 together with TP53 signaling (Table S4).
Our study also corroborates an expanding body of evidence showing that RUNX2, a transcription factor central to the control of osteoblast differentiation during skeletal development and remodelling, is frequently expressed at high levels in osteosarcoma biopsies , , , . The deregulation of RUNX2 in cancer has been linked to signalling pathways disrupted in tumorigenesis, including RB1 and TP53 , . RUNX2 is growth suppressive in normal osteoblasts, but can induce proliferation-specific genes if over-expressed . Loss of TP53 increases protein levels of RUNX2 in osteosarcoma cells by post-transcriptional mechanisms Loss of TP53 prevents post-expression of microRNA miR-34c, which directly targets RUNX2 . However, our study indicates that transcriptional mechanisms may also play an important role.
The RUNX2 gene is controlled by two promoters. The P1 promoter is activated at late stages of osteoblast differentiation to elevate RUNX2 levels in support of osteoblast maturation . Strikingly, we noted higher levels of expression of the RUNX2 P2 transcript in tumors which responded poorly to chemotherapy, closely replicating a previous finding from our lab . The P2 promoter contains multiple CpG doublets, and hypomethylation of this promoter may permit expression in mesenchymal stem cells . In addition, RUNX2 expressed from the P2 promoter regulates hypertrophy and proliferation of both chondrocytes and the closely-related immature osteoblasts –. The human osteosarcoma cell line SAOS-2 has persistently high RUNX2 protein levels, driven by the P2 promoter . Hence, preferential expression of RUNX2 from the ‘early-activated’ P2 promoter rather than the ‘late-activated’ P1 promoter in osteosarcomas suggests that osteosarcomas may originate from immature mesenchymal progenitor cells.
The samples in this study consisted of tumor resections both from patients who had been treated with chemotherapy and from pre-chemotherapeutic biopsies. Statistically significant gene expression differences between resections and biopsies existed for the expression levels of CDKN1A, MDM2, BCL2L1, PTEN, and WWOX. CDKN1A is activated in response to activation of the ATM/TP53 DNA damage checkpoint that accommodates double-stranded DNA repair and inhibits cell cycle progression by CDK4 , MDM2 is an inhibitor of TP53 , BCL2L1 is an anti-apoptotic factor , PTEN is a tumor suppressor commonly lost in osteosarcomas , and WWOX is a tumor suppressor that inhibits RUNX2 activity . Furthermore, all of these genes encode proteins important for cell cycle regulation. The IPA analysis of the data confirms significant relationships between proteins in theTP53-RB1-centred network of protein interactions that also involve PI3K, PTEN/Akt, MYC, and RECQL4 (Figure 3A). All of the aforementioned genes were more highly expressed in the resections. Our results are consistent with experiments in osteosarcoma cell lines that have shown that drug treatment induces growth arrest and increases levels of CDKN1A, MDM2, and BCL2L1 , .
In conclusion, our results provide preliminary evidence that RUNX2, CDC5L, MDM2, RECQL4, and CDK4 should be further investigated to determine their roles as predictive biomarkers in osteosarcoma, because collectively, their differential expression correlates with poor response to chemotherapy in our cohort.
Cluster map constructed using Cluster 3.0, showing the differential expression of the 16 gene set in osteosarcoma biopsy and resection cases. Sample numbers 1–36 were analyzed in triplicates whereas, samples 41–116 were analyzed in duplicates. Details of samples are presented in Table S2B. The seven replicates of the three osteosarcoma cell lines and the pooled human osteoblast control used in this comparison cluster to the right as four distinct groupings (C–R). From left to right they are SAOS samples CBAAMLB; MG63 samples GIHQGHP; human osteoblast control samples EDFDEON; and U2OS samples KJLJKSR.
Expression changes for each of the 32 patient samples were normalized, averaged, and then ratios relative to normal human osteoblast control were calculated. Statistically significant up-regulated expression in tumors relative to human osteoblasts was detected for CDKN1C, FOS, MYC, RECQL4, RUNX2, SPARC, SPP1, and WWOX. Down-regulation was observed for CDKN1A and TP53.
Codeset for expression analysis.
Table S2A: All nanoString data with normalization applied. Table S2B: Patient nanoString data sets with normalization applied. Patient samples in the first set were analyzed in triplicate, whereas samples in the second and third sets were analyzed in duplicate.
Gene expression levels differentiated groups of tumors. The unpaired two-tailed Student’s t-test was applied to derive genes that had significant (p<0.05) differential expression between the group with poor prognosis (n = 25) and the group with good prognosis (n = 15).
Previously published results show similar pathways of expression deregulation in osteosarcoma. Genes deregulated in osteosarcoma as identified by Kelly et al. (Additional file 14 Table S5 of reference 59) were functionally related to our experimental genes.
The authors would like to thank Kirsteen Maclean of nanoString technologies for her invaluable help with the cluster analysis and Jennifer Good from Queen’s University for her dedication and technical help.
Conceived and designed the experiments: MZ JAS AJvW. Performed the experiments: JWM SC-M. Analyzed the data: JWM MK MZ JAS. Contributed reagents/materials/analysis tools: MZ SC-M JAS. Wrote the paper: JWM SC-M MK AJvW MZ JAS.
- 1. Bielack SS, Kempf-Bielack B, Delling G, Exner GU, Flege S, et al. (2002) Prognostic factors in high-grade osteosarcoma of the extremities or trunk: an analysis of 1,702 patients treated on neoadjuvant cooperative osteosarcoma study group protocols. J Clin Oncol 20(3): 776–790.
- 2. Cleton-Jansen AM, Anninga JK, Briaire-de Bruijn IH, Romeo S, Oosting J, et al. (2009) Profiling of high-grade central osteosarcoma and its putative progenitor cells identifies tumorigenic pathways. Br J Cancer 101(11): 1909–1918.
- 3. Martin JW, Squire JA, Zielenska M (2012) The genetics of osteosarcoma. Sarcoma 2012: 627254.
- 4. Broadhead ML, Clark JC, Myers DE, Dass CR, Choong PF (2011) The molecular pathogenesis of osteosarcoma: a review. Sarcoma 2011: 959248.
- 5. Selvarajah S, Yoshimoto M, Ludkovski O, Park PC, Bayani J, et al. (2008) Genomic signatures of chromosomal instability and osteosarcoma progression detected by high resolution array CGH and interphase FISH. Cytogenet Genome Res 122(1): 5–15.
- 6. Chibon F, Lagarde P, Salas S, Perot G, Brouste V, et al. (2010) Validated prediction of clinical outcome in sarcomas and multiple types of cancer on the basis of a gene expression signature related to genome complexity. Nat Med 16(7): 781–787.
- 7. Helman LJ, Meltzer P (2003) Mechanisms of sarcoma development. Nat Rev Cancer 3(9): 685–694.
- 8. Scott MC, Sarver AL, Gavin KJ, Thayanithy V, Getzy DM, et al. (2011) Molecular subtypes of osteosarcoma identified by reducing tumor heterogeneity through an interspecies comparative approach. Bone 49(3): 356–367.
- 9. Lu XY, Lu Y, Zhao YJ, Jaeweon K, Kang J, et al. (2008) Cell cycle regulator gene CDC5L, a potential target for 6p12-p21 amplicon in osteosarcoma. Mol Cancer Res 6(6): 937–946.
- 10. Sadikovic B, Thorner P, Chilton-Macneill S, Martin JW, Cervigne NK, et al. (2010) Expression analysis of genes associated with human osteosarcoma tumors shows correlation of RUNX2 overexpression with poor response to chemotherapy. BMC Cancer 10: 202.
- 11. Sadikovic B, Yoshimoto M, Chilton-MacNeill S, Thorner P, Squire JA, et al. (2009) Identification of interactive networks of gene expression associated with osteosarcoma oncogenesis by integrated molecular profiling. Hum Mol Genet 18(11): 1962–1975.
- 12. Namlos HM, Kresse SH, Muller CR, Henriksen J, Holdhus R, et al. (2012) Global gene expression profiling of human osteosarcomas reveals metastasis-associated chemokine pattern. Sarcoma 2012: 639038.
- 13. Northcott PA, Shih DJ, Remke M, Cho YJ, Kool M, et al. (2012) Rapid, reliable, and reproducible molecular sub-grouping of clinical medulloblastoma samples. Acta Neuropathol 123(4): 615–626.
- 14. Geiss GK, Bumgarner RE, Birditt B, Dahl T, Dowidar N, et al. (2008) Direct multiplexed measurement of gene expression with color-coded probe pairs. Nat Biotechnol 26(3): 317–325.
- 15. Nathan SS, Pereira BP, Zhou YF, Gupta A, Dombrowski C, et al. (2009) Elevated expression of Runx2 as a key parameter in the etiology of osteosarcoma. Mol Biol Rep 36(1): 153–158.
- 16. Xiao ZS, Thomas R, Hinson TK, Quarles LD (1998) Genomic structure and isoform expression of the mouse, rat and human Cbfa1/Osf2 transcription factor. Gene 214(1–2): 187–197.
- 17. Dalla-Torre CA, Yoshimoto M, Lee CH, Joshua AM, de ToledoSR, et al. (2006) Effects of THBS3, SPARC and SPP1 expression on biological behavior and survival in patients with osteosarcoma. BMC Cancer 6: 237.
- 18. Gamberi G, Benassi MS, Bohling T, Ragazzini P, Molendini L, et al. (1998) C-myc and c-fos in human osteosarcoma: prognostic value of mRNA and protein expression. Oncology 55(6): 556–563.
- 19. Shimizu T, Ishikawa T, Sugihara E, Kuninaka S, Miyamoto T, et al. (2010) c-MYC overexpression with loss of Ink4a/Arf transforms bone marrow stromal cells into osteosarcoma accompanied by loss of adipogenesis. Oncogene 29(42): 5687–5699.
- 20. Wang ZX, Yang JS, Pan X, Wang JR, Li J, et al. (2010) Functional and biological analysis of Bcl-xL expression in human osteosarcoma. Bone 47(2): 445–454.
- 21. Yoshida A, Ushiku T, Motoi T, Beppu Y, Fukayama M, et al. (2012) MDM2 and CDK4 immunohistochemical coexpression in high-grade osteosarcoma: correlation with a dedifferentiated subtype. Am J Surg Pathol 36(3): 423–431.
- 22. Feugeas O, Guriec N, Babin-Boilletot A, Marcellin L, Simon P, et al. (1996) Loss of heterozygosity of the RB gene is a poor prognostic factor in patients with osteosarcoma. J Clin Oncol 14(2): 467–472.
- 23. Freeman SS, Allen SW, Ganti R, Wu J, Ma J, et al. (2008) Copy number gains in EGFR and copy number losses in PTEN are common events in osteosarcoma tumors. Cancer 113(6): 1453–1461.
- 24. Kurek KC, Del Mare S, Salah Z, Abdeen S, Sadiq H, et al. (2010) Frequent attenuation of the WWOX tumor suppressor in osteosarcoma is associated with increased tumorigenicity and aberrant RUNX2 expression. Cancer Res 70(13): 5577–5586.
- 25. Mohseny AB, Tieken C, van der Velden PA, Szuhai K, de Andrea C, et al. (2010) Small deletions but not methylation underlie CDKN2A/p16 loss of expression in conventional osteosarcoma. Genes Chromosom Cancer 49(12): 1095–1103.
- 26. Tsuchiya T, Sekine K, Hinohara S, Namiki T, Nobori T, et al. (2000) Analysis of the p16INK4, p14ARF, p15, TP53, and MDM2 genes and their prognostic implications in osteosarcoma and Ewing sarcoma. Cancer Genet Cytogenet 120(2): 91–98.
- 27. Wunder JS, Gokgoz N, Parkes R, Bull SB, Eskandarian S, et al. (2005) TP53 mutations and outcome in osteosarcoma: a prospective, multicenter study. J Clin Oncol 23(7): 1483–1490.
- 28. Lapenna S, Giordano A (2009) Cell cycle kinases as therapeutic targets for cancer. Nature Rev 8(7): 547–566.
- 29. Janssens N, Janicot M, Perera T, Bakker A (2004) Housekeeping genes as internal standards in cancer research. Mol Diagn 8(2): 107–113.
- 30. Maire G, Yoshimoto M, Chilton-MacNeill S, Thorner PS, Zielenska M, et al. (2009) Recurrent RECQL4 imbalance and increased gene expression levels are associated with structural chromosomal instability in sporadic osteosarcoma. Neoplasia 11(3): 260–268.
- 31. Teplyuk NM, Galindo M, Teplyuk VI, Pratap J, Young DW, et al. (2008) RUNX2 regulates G protein-coupled signaling pathways to control growth of osteoblast progenitors. J Biol Chem 283(41): 27585–27597.
- 32. Kuijjer ML, Rydbeck H, Kresse SH, Buddingh EP, Lid AB, et al. (2012) Identification of osteosarcoma driver genes by integrative analysis of copy number and gene expression data. Genes Chromosom Cancer 51(7): 696–706.
- 33. Clark JC, Dass CR, Choong PF (2008) A review of clinical and molecular prognostic factors in osteosarcoma. J Cancer Res Clin Oncol 134(3): 281–297.
- 34. Malkov VA, Serikawa KA, Balantac N, Watters J, Geiss G, et al. (2009) Multiplexed measurements of gene signatures in different analytes using the Nanostring nCounter Assay System. BMC Res Notes 2: 80.
- 35. Turner NC, Reis-Filho JS (2012) Genetic heterogeneity and cancer drug resistance. Lancet Oncol 13(4): e178–185.
- 36. Thangavel S, Mendoza-Maldonado R, Tissino E, Sidorova JM, Yin J, et al. (2010) Human RECQ1 and RECQ4 helicases play distinct roles in DNA replication initiation. Cell Mol Biol 30(6): 1382–1396.
- 37. Wang LL, Gannavarapu A, Kozinetz CA, Levy ML, Lewis RA, et al. (2003) Association between osteosarcoma and deleterious mutations in the RECQL4 gene in Rothmund-Thomson syndrome. J Natl Cancer Inst 95(9): 669–674.
- 38. Bernstein HS, Coughlin SR (1998) A mammalian homolog of fission yeast Cdc5 regulates G2 progression and mitotic entry. J Biol Chem 273(8): 4666–4671.
- 39. Zhang N, Kaur R, Akhter S, Legerski RJ (2009) Cdc5L interacts with ATR and is required for the S-phase cell-cycle checkpoint. EMBO reports 10(9): 1029–1035.
- 40. Malumbres M, Barbacid M (2007) Cell cycle kinases in cancer. Curr Opin Genet Dev 17(1): 60–65.
- 41. Dujardin F, Binh MB, Bouvier C, Gomez-Brouchet A, Larousserie F, et al. (2011) MDM2 and CDK4 immunohistochemistry is a valuable tool in the differential diagnosis of low-grade osteosarcomas and other primary fibro-osseous lesions of the bone. Mod Pathol 24(5): 624–637.
- 42. Kyriazoglou AI, Vieira J, Dimitriadis E, Arnogiannaki N, Teixeira MR, et al. (2012) 12q amplification defines a subtype of extraskeletal osteosarcoma with good prognosis that is the soft tissue homologue of parosteal osteosarcoma. Cancer Genet 205(6): 332–336.
- 43. Smida J, Baumhoer D, Rosemann M, Walch A, Bielack S, et al. (2010) Genomic alterations and allelic imbalances are strong prognostic predictors in osteosarcoma. Clin Cancer Res. 16(16): 4256–4267.
- 44. Won KY, Park HR, Park YK (2009) Prognostic implication of immunohistochemical Runx2 expression in osteosarcoma. Tumori 95(3): 311–316.
- 45. Lee JS, Thomas DM, Gutierrez G, Carty SA, Yanagawa S, et al. (2006) HES1 cooperates with pRb to activate RUNX2-dependent transcription. J Bone Miner Res 21(6): 921–933.
- 46. Lengner CJ, Steinman HA, Gagnon J, Smith TW, Henderson JE, et al. (2006) Osteoblast differentiation and skeletal development are regulated by Mdm2-p53 signaling. J Cell Biol 172(6): 909–921.
- 47. van der Deen M, Taipaleenmaki H, Zhang Y, Teplyuk NM, Gupta A, et al. (2013) MicroRNA-34c inversely couples the biological functions of the Runt-related transcription factor RUNX2 and the tumor suppressor p53 in osteosarcoma. J Biol Chem 288(29): 21307–21319.
- 48. Liu JC, Lengner CJ, Gaur T, Lou Y, Hussain S, et al. (2011) Runx2 protein expression utilizes the Runx2 P1 promoter to establish osteoprogenitor cell number for normal bone formation. J Biol Chem 286(34): 30057–30070.
- 49. Kang MI, Kim HS, Jung YC, Kim YH, Hong SJ, et al. (2007) Transitional CpG methylation between promoters and retroelements of tissue-specific genes during human mesenchymal cell differentiation. J Cell Biochem 102(1): 224–239.
- 50. Lucero CM, Vega OA, Osorio MM, Tapia JC, Antonelli M, et al. (2013) The cancer-related transcription factor Runx2 modulates cell proliferation in human osteosarcoma cell lines. J Cell Physiol 228(4): 714–723.
- 51. Stein GS, Lian JB, van Wijnen AJ, Stein JL, Montecino M, et al. (2004) Runx2 control of organization, assembly and activity of the regulatory machinery for skeletal gene expression. Oncogene 23(24): 4315–4329.
- 52. Thomas DM, Johnson SA, Sims NA, Trivett MK, Slavin JL, et al. (2004) Terminal osteoblast differentiation, mediated by runx2 and p27KIP1, is disrupted in osteosarcoma. J Cell Biol 167(5): 925–934.
- 53. Terry A, Kilbey A, Vaillant F, Stewart M, Jenkins A, et al. (2004) Conservation and expression of an alternative 3′ exon of Runx2 encoding a novel proline-rich C-terminal domain. Gene 336(1): 115–125.
- 54. Mauro M, Rego MA, Boisvert RA, Esashi F, Cavallo F, et al. (2012) p21 promotes error-free replication-coupled DNA double-strand break repair. Nucleic Acids Res 40(17): 8348–8360.
- 55. Haupt Y, Maya R, Kazaz A, Oren M (1997) Mdm2 promotes the rapid degradation of p53. Nature 387(6630): 296–299.
- 56. Brunelle JK, Letai A (2009) Control of mitochondrial apoptosis by the Bcl-2 family. J Cell Sci 122(Pt 4): 437–441.
- 57. Gallaher BW, Berthold A, Klammt J, Knupfer M, Kratzsch J, et al. (2000) Expression of apoptosis and cell cycle related genes in proliferating and colcemid arrested cells of divergent lineage. Cell Mol Biol 46(1): 79–88.
- 58. Sato N, Mizumoto K, Maehara N, Kusumoto M, Nishio S, et al. (2000) Enhancement of drug-induced apoptosis by antisense oligodeoxynucleotides targeted against Mdm2 and p21WAF1/CIP1. Anticancer Res 20(2A): 837–842.
- 59. Kelly AD, Haibe-Kains B, Janeway KA, Hill KE, Howe E, et al. (2013) MicroRNA paraffin-based studies in osteosarcoma reveal reproducible independent prognostic profiles at 14q32. Genome Medicine 5(2): 1–12.
- 60. Martin JW, Zielenska M, Stein GS, van Wijnen AJ, Squire JA (2011) The Role of RUNX2 in osteosarcoma oncogenesis. Sarcoma 2011: 282745.