CROSS-REFERENCE TO RELATED APPLICATION This application claims priority to U.S. Provisional Patent Application Serial No. 61/724,509, filed November 9, 2012, which is incorporated herein by reference. SUMMARY

This disclosure describes, in one aspect, a method that generally includes analyzing a tissue sample from a subject for expression of a 14q32 miRNA, wherein a decrease in expression of the 14q32 miRNA compared to a normal control indicates that the subject has or is at risk of having osteosarcoma.

In some embodiments, expression of the 14q32 miRNA of no more than 10% of a normal control further indicates that the subject has or is at risk of having an aggressive osteosarcoma.

In some embodiments, the 14q32 miRNA can include miR-382, miR-134, or miR- 544. In other embodiments, the 14q32 miRNA can include miR-379, miR-1197, miR-329-1, miR-494, miR-495, miR-376C, miR-1185-1, miR-1185-2, miR-381, miR-655, miR-487A, miR-382, miR-496, miR-154, miR-369, miR-127, miR-299, miR-665, miR-770, miR-432, miR-668, miR-539, miR-433, miR-323, miR-496, miR-654, miR-409, miR-380, miR-485, miR-136, miR-376, miR-431, miR-370, miR-889, miR-381, miR-299, miR-379, miR-411, miR-655, miR-432, miR-487, miR-410, miR-493, miR-656, miR-485, miR-377 or miR-541.

The above summary of the present invention is not intended to describe each disclosed embodiment or every implementation of the present invention. The description that follows more particularly exemplifies illustrative embodiments. In several places throughout the application, guidance is provided through lists of examples, which examples can be used in various combinations. In each instance, the recited list serves only as a representative group and should not be interpreted as an exclusive list. BRIEF DESCRIPTION OF THE FIGURES

FIG. 1. 14q32 miRNA expression levels and metastasis in human osteosarcoma. (A) Correlation network of 14q32 miRNAs expression in human osteosarcoma samples.

Correlation network was constructed with Pearson correlation R >0.9 as the edges. The resulting correlation sub-network contained thirty-six miRNAs and 33 of them mapped to the 14q32 locus. (B) Microarray based transcript levels of highly correlated 14q32 miRNAs in osteosarcoma tumors shown relative to the level observed in FT- 14 (human osteosarcoma primary tumor). miRNA profiling data from human osteosarcoma tumors where metastasis was observed show stronger downregulation of 14q32 miRNAs (highlighted) than primary osteosarcoma tumors where metastasis was not observed. (C) Microarray based transcript levels for miR-382 and miR-154. miR-382 and miR-154 show a high level of correlation (R = 0.95). (D) qRT-PCR validation for miR-382 and miR-154 expression levels obtained in primary osteosarcoma tumors (FT- 14, -7) where metastasis was not observed and three primary osteosarcoma tumors with low levels of (FT- 18, -13, -12) where metastasis was observed. miR-382 and miR-154 show a high level of correlation (R = 0.99). Measurements were carried out in triplicate and were normalized to the expression levels in two independent normal bone tissues that were also carried out in triplicate.

FIG. 2. Prognostic significance of genes correlated with 14q32 miRNAs in canine osteosarcoma. (A) Unsupervised clustering of human mRNAs (n=385) that shows high-level direct correlations to miR-382 for two normal bone tissues and 11 osteosarcoma tumors, for which we have both mRNA profiles and miRNA profiles. A positive correlation to miR-382 is shown in lighter grey (n=288), and a negative correlation to miR-382 is shown in darker grey (n=97), at the right of the heatmap. Directional functional enrichment analyses

(Ingenuity Pathways Analyses) shows that correlating transcripts are enriched in transcripts involved in metastasis, and direction of change is consistent with increased activity of metastatic function in tumors with decreased levels of 14q32 miRNA member, miR-382. Identities of positive and negatively correlating mRNA are provided in Table 6 and 7. (B) Unsupervised hierarchical clustering heatmap of canine mRNA found in 26 canine OS- derived samples that correspond to human mRNAs correlating to miR-382. The bars to at the right side of each heatmap represent correlations to miR-382 (positive = lighter grey or negative= darker grey) observed in the human data indicating that the direction of change shows similar trends between human and canine data. (C) Kaplan-Meier survival curves generated using the groups shown in FIG. 2B. Survival times are significantly different between the two groups of tumors (p-value = 0.02).

FIG. 3. 14q32 miRNA expression level, metastasis, and outcome in human OS. (A) miR-382 expression levels determined by qRT-PCR in osteosarcoma primary tumor samples with clinical follow-up information. The miRNA expression levels were normalized relative to normal bone tissues. Osteosarcoma samples were ranked based on expression levels of miR-382 from highest to lowest. Primary osteosarcoma patient samples that later developed metastases are highlighted and samples from patients who died due to disease are shown in black. Samples were grouped based on whether they were above or below the median expression level. (B) Kaplan-Meier analysis of metastasis in osteosarcoma patients based on miR-382 expression. Group 1 and Group 2 are significantly different (p-value = 0.01).

Patients with lowest levels of miR-382 expression showed increased likelihood of metastasis. (C) Kaplan-Meier analysis of survival in osteosarcoma patients based on miR-382 expression. Patients with lowest levels of miR-382 expression showed decreased likelihood of survival.

FIG. 4. 14q32 orthologous region transcript levels and outcome in canine

osteosarcoma. (A) miR-134 and (B) miR-544 expression levels determined by qRT-PCR in canine osteosarcoma primary tumor samples with clinical follow-up information. The miRNA expression levels were normalized relative to reactive osteoblasts. Osteosarcoma samples were ranked based on expression levels of miR-134 from highest to lowest. Samples were grouped based on whether they were above or below the median expression levels. (C and D) Kaplan-Meier analysis of survival in osteosarcoma patients based on miR-134 and miR-544 expression. Survival of dogs with osteosarcoma in Group 1 and Group 2 are significantly different (miR-134 p-value = 0.004, miR-544 p-value = 0.01). Dogs with lowest levels of miR-134 or miR-544 expression showed decreased likelihood of survival.

FIG. 5. Homology between human 14q32 miRNAs and CFA 8:72.3 Mb miRNAs locus. Data was obtained from Entrez gene database. (A-B) Conservation at the gene level. (C-D) Conservation at the miRNA level. While the order is highly maintained only certain miRNAs are shown. The position of miR-379 was used to orient the orthologous canine region of the genome to the human 14q32 locus. DETAILED DESCRIPTION OF ILLUSTRATIVE EMBODIMENTS

Deregulation of microRNA (miRNA) transcript levels has been observed in many types of tumors including, for example, osteosarcoma. Molecular pathways regulated by differentially expressed miRNAs may contribute to the heterogeneous tumor behaviors observed in naturally occurring cancers. Thus, tumor-associated miRNA expression may provide informative biomarkers for disease outcome and metastatic potential in osteosarcoma patients.

We have discovered an inverse correlation between aggressive tumor behavior— e.g., increased metastatic potential and accelerated time to death— and the residual expression of 14q32 miRNAs (using miR-382 as a representative miRNA of the 14q32 miRNA family) in a series of clinically annotated samples from human osteosarcoma patients. We also have discovered a comparable decrease in expression of orthologous 14q32 miRNAs in canine osteosarcoma samples, with conservation of the inverse correlation between aggressive behavior and expression of orthologous miRNA miR- 134 and miR-544. Thus,

downregulation of 14q32 miRNA expression is an evolutionarily conserved mechanism that contributes to the biological behavior of osteosarcoma. Consequently, quantifying representative transcripts of the 14q32 miRNA family such as, for example, miR-382, miR- 134, and/or miR-544, provide prognostic and/or predictive markers that can assist in the management of patients with this disease.

As used herein, the term "14q32 miRNA" refers to a microRNA transcribed from any portion of the 14q32 locus of the human genome and/or a microRNA transcribed from an orthologous locus of a chromosome of another species such as, for example, the CFA 8 locus of the canine genome. For example, a particular miRNA that is transcribed from the 14q32 locus of the human genome may be characterized as a human 14q32 miRNA, while a particular miRNA that is transcribed from the CFA 8 locus of the canine genome may be characterized as a canine 14q32 miRNA.

Osteosarcoma is a common primary bone tumor, often affecting children and adolescents (Mirabello et al, Cancer 2009, 115: 1531-1543; Kansara et al, DNA Cell Biol 2007, 26: 1-18). The molecular understanding of osteosarcoma has progressed slowly compared to other cancer types due, at least in part, to its infrequent occurrence and/or the heterogeneous nature of osteosarcoma in humans. The occurrence of an orthologous disease in dogs, however, provides an opportunity to surmount some of these challenges.

Osteosarcoma is more common in dogs (estimated at > 10,000 cases per year), with exquisite breed predilection. Osteosarcoma arises spontaneously in dogs and its biology is similar to the human disease both anatomically and functionally (Khanna et al., Nat

Biotechnol 2006, 24: 1065-1066). The chronology of osteosarcoma is adapted to lifetime, so progression of osteosarcoma in dogs takes place in a time frame (~2 years) that allows rapid assessment of risk, prediction, and outcome (Khanna et al., Nat Biotechnol 2006, 24: 1065- 1066). The availability of detailed genome maps for humans and dogs, and the extensive homology in sequence and gene order between both species makes integration and

extrapolation of genomic data feasible (Lindblad-Toh et al., Nature 2005, 438:803-819;

Shearin et al, Dis Model Mech 2010, 3:27-34).

We have previously identified sets of differentially expressed genes that are associated with tumor behavior in both canine and human osteosarcoma (Scott et al., Bone 2011, 49:356- 367). Specifically, we characterized gene signatures that can stratify canine and human osteosarcoma according to the predicted biological behavior of the tumors (Scott et al., Bone 2011, 49:356-367). The signatures in dogs with the worst outcomes are characterized by increased cell cycle gene expression that regulates G2/M transition and DNA damage genes. In contrast, the signatures that are typified in dogs with better outcomes include increased expression of "microenvironment interactions" genes. These patterns also were observed in human osteosarcoma tumor datasets, but the mechanisms responsible for generating these signatures have not yet been fully defined.

Over 40 miRNAs are present between the imprinted genes DLK1 and DI03 at the human 14q32 locus (Seitz et al, Genome Res 2004, 14: 1741-1748). We generated a correlation-based network using miRNA expression levels determined by array analyses and observed decreases in expression levels of 14q32 miRNA that are highly correlated in each of the individual tumor tissue samples. The analysis further revealed that expression levels between various of the miRNAs found at the 14q32 locus were highly correlated (FIG. 1 A; Table 8). The high correlation in expression levels between 14q32 miRNAs suggests that an individual miRNA from this locus could represent the expression levels of most of 14q32 miRNAs in osteosarcoma. To demonstrate that the high level of correlation observed for 14q32 miRNAs was derived from meaningful decreases in expression levels, we plotted the expression levels of the 14q32 miRNA members in human osteosarcoma relative to the human osteosarcoma primary tumor sample FT- 14. Several of the osteosarcoma tumor samples showed 14q32 miRNA levels similar to those observed in FT- 14 while several of the other osteosarcoma tumors showed much more sizable decreases in 14q32 miRNAs. The magnitude of miRNA expression decrease ranged from approximately 3-fold to approximately 20-fold relative to the levels observed in normal bone tissue by miRNA microarray analyses (Sarver et al., Laboratory Invest 2010, 90:753-761; Thayanithy et al, Bone 2012, 50: 171-181). The heterogeneity in expression levels was seen in 14q32 locus miRNAs, and is specifically shown for two members at this locus: miR-382 and miR-154 (FIG. 1C). The expression levels observed for miR-382 and miR-154 showed a high level of correlation (R = 0.95), exemplary of the high correlations observed across the 14q32 miRNAs. In order to confirm that the results obtained by miRNA microarray for the magnitude of the 14q32 miRNA decreases were meaningful, we carried out qRT-PCR for miR-382 and miR-154. The results show that, relative to normal bone, osteosarcoma samples FT- 14 and FT-7 exhibited an approximately 3- fold decrease while osteosarcoma samples FT-13, FT- 18, and FT- 12 showed an

approximately 10-fold to approximately 20-fold decrease in the expression levels of these two 14q32 miRNAs (FIG. ID). The qRT-PCR results for 14q32 miRNA correlation also showed high correlation (R 2 = 0.99) similar to the correlation observed by microarray (R 2 =0.95).

To further understand the role of the 14q32 miRNAs in human osteosarcoma, we profiled the mRNA transcript levels of tumor samples for which we also had miRNA profiles. We then identified mRNA that positively and negatively correlated to miR-382. Since all of the tested 14q32 miRNAs expression levels were highly correlated on a sample-by-sample basis, we selected miR-382 as a representative of 14q32 miRNAs. The results of this analysis in human osteosarcoma samples uncovered 288 genes that were positively (see Table 6) correlated to miR-382 expression and 97 genes that were negatively (see Table 7) correlated to miR-382 expression (R > 0.7). All of these 385 genes and their direction of change in relation to miR-382 are shown as a heatmap in FIG. 2A. Ingenuity Pathway Analysis (IP A) (accessed April 14, 2012) showed that genes involved in the functional category "metastasis" were enriched in this signature (Enrichment p-value < 0.001 after Benjamini Hochberg multiple testing correction). Furthermore, genes of this signature in the functional category "metastasis" were predicted to have increased activity (Regulation Z-score 2.123) in osteosarcoma tumor samples that showed the lowest levels of 14q32 miRNAs on the basis of the direction of change of the correlated genes. This provides a potential molecular rationale for the observation of metastasis being present in the samples with the largest decreases in 14q32 miRNA transcript levels.

As a next step in comparative assessment, we mapped the human osteosarcoma genes that positively and negatively correlated with miR-382 (FIG. 2A, direction of change marked in lighter and darker greys) to orthologous canine genes that were included in the Affymetrix canine_2.0 platform (FIG. 2B). Mapping of the human to canine genes provided 219 canine genes, which we then used to perform unsupervised hierarchical clustering of the 26 canine osteosarcoma samples. These 219 genes segregated the canine osteosarcoma samples into two distinct branches. Kaplan-Meier analysis using death as an outcome revealed a statistically significant (p< 0.05) difference in survival between dogs in these two distinct branches (FIG. 2C). Dogs in which miR-382-positively correlated mRNAs were expressed at lower levels and in which miR-382-negatively correlated mRNAs were expressed at higher levels had significantly shorter survival times than the other group of dogs. This result shows that genes that correlate with miR-382 are capable of showing survival trends in dogs with osteosarcoma that are consistent with our analysis in humans, thus independently supporting a prognostic role for the 14q32 miRNA transcript level in osteosarcoma.

Because metastasis genes are enriched in miR-382-correlated gene signatures in canine osteosarcoma and miR-382 expression levels showed potential prognostic significance (e.g., ability to predict poor outcome) in the canine osteosarcoma cohort, we tested whether expression levels of 14q32 miRNAs in human showed similar association with outcomes. We analyzed the miRNA profiles obtained from the initial cohort of human osteosarcoma tumor tissue samples. Analysis revealed an inverse relationship: the osteosarcoma samples with the lowest levels of 14q32 miRNAs were from patients that developed metastatic tumors (FIG. 1B-D and 2A). This human osteosarcoma sample cohort was incompletely annotated (no survival data available), but a Mann Whitney test showed that a potential association was present between levels of 14q32 expression levels and metastasis (p = 0.1357). To further examine the association between low 14q32 miRNA levels in osteosarcoma tumors and clinical outcomes, we tested a representative 14q32 miRNA, miR-382, in an independent set of primary tumor samples from 16 human osteosarcoma patients that had more robust clinical follow-up annotation (see Table 4). As determined by qRT-PCR, all of the osteosarcoma tumor tissue samples in this independent set showed lower levels of miR- 382 compared to normal bone tissues (FIG. 3A). As we noticed in the first human

osteosarcoma dataset, the lowest levels of miR-382 were found in primary osteosarcoma tumors from patients where metastases and/or death were later observed as follow-up. We further noticed that there was a negative correlation between the levels of miR-382 expression and patient survival (FIG. 3A; Table 4). To test whether lower levels of 14q32 were significantly associated with metastasis, we used the Mann Whitney test. This test allowed us to reject the null hypothesis that there was no association between 14q32 level and metastasis (p < 0.01).

Further, to examine the prognostic utility of the 14q32 miRNA level, we split the cohort into two groups based on whether the miR-382 level was above the median value

(Group 1) or below the median value (Group 2). We compared the two groups using Kaplan- Meier analyses using both metastasis and death as outcomes (FIG. 3B and 3C). Human osteosarcoma tumor samples that had the lowest levels of miR-382 were at higher risk for metastasis (p = 0.01) and death (p = 0.08) than patients with moderate miR-382 expression independently confirming the observation that 14q32 miRNAs are a potential prognostic tool for osteosarcoma.

We examined the orthologous 14q32 locus miRNAs and genes within the canine genome using the Entrez gene database and were able to find them on CFA 8 (FIG. 5).

Similar to human 14q32 locus, canine orthologous miRNAs were also found in the syntenic region between genes DLK1 and DI03. The order of the miRNA is retained in the canine genome although differences in spacing were observed between human and canine versions.

The orthologous canine miRNAs on CFA 8 can, in a manner similar to that observed for

14q32 miRNAs in humans, show decreases in transcript level in canine osteosarcoma tumors.

Moreover, the magnitude of these decrease in expression can provide prognostic utility in canines. For example,, we determined the levels of miR-544 and miR-134 (based on identical mature miRNA sequences) in 16 canine osteosarcoma tumors from which RNA was available by qRT-PCR (FIG. 4A). In all 16 cases, expression decreases were observed in the miRNA relative to reactive canine osteoblasts. As observed in human osteosarcoma, the levels of miR- 544 and miR-134 were directly correlated (R ² = 0.88)

The canine osteosarcoma samples that were analyzed by qRT-PCR were derived from dogs with overall survival times ranging from two to 606 days with a median survival of

131.5 days (Table 5). The canine cohort was separated into two groups using the same method used for the human tumor samples. Group 1 was composed of tumor samples with miR-134 (FIG. 4A) and miR-544 (FIG. 4B) expression levels above the median survival days and Group 2 was composed of tumor samples below the median, and these two groups were interrogated via Kaplan-Meier survival analysis (FIG. 4D). Dogs with tumors with more severely reduced levels of miR-134 had higher risk for death (p= 0.004) than dogs with tumors with more moderate decreases in miR-134 expression (FIG. 4C). Similar results were obtained using miR-544 expression levels (p= 0.012, FIG. 4D). Together, these results indicate that the 14q32 miRNA expression levels have prognostic utility in osteosarcoma, a disease that shows conserved features between humans and dogs.

Molecular markers to identify osteosarcoma patients at risk for metastatic progression are important for guiding decisions regarding suitable treatment options. Recently, several molecular markers based on gene expression in tumor tissues have been identified in osteosarcoma. Higher levels of oncogene YY1 expression in primary sites of osteosarcoma are associated with metastasis and poor clinical outcomes (de Nigris et al., BMC Cancer 2011, 11 :472). Similarly, overexpression of markers such as cancer-testis antigens, Cystein Rich Protein (CYR61) and Melanoma Antigen Family A (MAGE A) are implicated in predicting tumor progression and prognosis in osteosarcoma (Zou et al, Cancer 2012, 118: 1845-1855, Sabile et al., J Bone Mineral Res 2012, 27:58-67). Additionally, the presence of tumor- infiltrating macrophages is associated with metastasis suppression in high-grade osteosarcoma (Buddingh et al, Clinical Cancer Res 2011, 17:21 10-2119). These studies show that the development of biomarkers for tumor progression and outcomes have clinical significance and may help develop patient specific treatment strategies.

The cluster of miRNAs at the 14q32 locus show decreased levels of expression in human osteosarcoma (Thayanithy et al., Bone 2012, 50: 171-181). Here, we report decreases in miRNA in the orthologous region of CFA 8 cluster members in canine osteosarcoma tumors. This is consistent with the presence of a commonly conserved mechanism of tumorigenesis involving the orthologous 14q32 locus associated genes and miRNAs between these species. Thus, the cross species comparison of human and canine expression profiles provides an opportunity to define the existence of evolutionarily conserved molecular signatures and to determine whether those signatures influence the outcome of our three patient cohorts (two human and one canine).

In both the human and canine osteosarcoma cases, the human 14q32 miRNAs and its canine orthologs appear to be regulated by a common mechanism, based on the high levels of expression correlation observed between the transcript levels of each member. Analyses of the SMED data reveal that 14q32 miRNA, as a group, are lost in a subset of GIST (3/14) and MFH (5/29). We conclude from this that the level of 14q32 miRNAs in general can be estimated from the measurement of a single member of the family and that measurement of the majority of the 14q32 miRNAs would provide highly similar data. Polycistronic regulation of these miRNAs has previously been proposed as a regulatory mechanism active at this locus and our data supports this concept (Seitz et al, Genome Res 2004, 14: 1741- 1748).

In an initial set of human osteosarcoma tumors miRNA microarray profiling data we observed that lower levels of 14q32 miRNA were associated with metastases. We further validated this by qRT-PCR for both miR-382 and miR-154 for osteosarcoma tumors from this initial dataset. We also showed that the lowest levels of miR-382 were significantly associated with metastasis and significantly worse survival in a second independent set of human primary osteosarcoma tumors where full clinical follow-up information was available. Taken together, these results suggest that the level of miR-382 may have prognostic utility for predicting the likelihood of developing metastatic tumors and disease outcomes. This observation is supported by three independent dataset from two different species. The association between reduced expression of 14q32 miRNAs and overall survival in humans represented a trend. In dogs, however, the magnitudes of miR-134 and miR-154 reduction were negative predictors of survival. This may be due, at least in part, to companion animal patients being treated with quality of life as the guiding principle. Therefore, metastatic disease is seldom treated aggressively and recurrence often leads to a decision of humane euthanasia, creating little difference between the disease-free interval and overall survival (Modiano et al, Molecular Therapy 2012, 20:2234-43).

Decreases in 14q32 miRNA levels stabilize cMYC expression in osteosarcoma and subsequently increase the expression of oncogenic miR-17-92 miRNAs (Thayanithy et al., Bone 2012, 50:171-181). Deregulations in 14q32 miRNA-cMYC network are also associated with increased proliferation and apoptotic escape in osteosarcoma (Thayanithy et al, Bone 2012, 50: 171-181). The prognostic relevance of decreasing 14q32 miRNA levels may be explained 1) by increased activation of the cMYC/miR- 17-92 network leading to poor outcomes 2) via regulation of an alternative cancer signaling pathway(s) either directly or indirectly or 3) as a biomarker whereby decreased levels of 14q32 miRNAs are observed in osteosarcoma cases with poor outcomes.

Genes involved in metastasis are significantly enriched in transcripts correlated to 14q32 miRNAs and the direction of change indicates that metastasis function is also significantly increased (see Tables 6 and 7). CDK5 and TWIST 1 can increase metastasis (Feldmann et al, Cancer Res 2010, 70:4460-4469; Reis et al, Oncogene 2004, 23:6684-6692) and are upregulated in osteosarcoma tumors with low levels of 14q32 miRNAs. IFNB1 (Kornbluth et al, J Leukocyte Biol 2006, 80: 1084-1102), TEK (Lin et al, Proc Natl Acad Sci USA 1998, 95:8829-8834) and COL18A1 (Brideau et al, Cancer Res 2007, 67: 11528-11535; Mendoza et al., Cancer Res 2004, 64:304-310) can decrease metastasis and are downregulated in tumors with low levels of 14q32 miRNAs. TK1 can be negatively correlated to 14q32 miRNA levels, is involved in the catalysis of phosphorylation of thymidine to

deoxythymidine monophosphate, expressed at high levels in proliferating cells, and appears to correlate with high risk in multiple cancer types (Konoplev et al., Am J Clin Pathol 2010, 134:472-477; Xu et al, Tumour Biol 2012, 33:475-483; He et al, Nucleosides, Nucleotides & Nucleic Acids 2010, 29:352-358). This suggests that genes correlated with 14q32 miRNA expression patterns may provide a molecular rational for the observed association between low levels of 14q32 and metastasis and poor survival outcomes and can be further

investigated as markers of prognosis.

We have previously demonstrated that comparative studies of human and canine osteosarcoma patients allow study of evolutionary conserved mechanisms of tumorigenesis associated with outcomes (Scott et al, Bone 2011, 49:356-367). A majority of the dogs evaluated in this study (21/25 in the correlating miRNA analysis and 12/16 in the qRT-PCR analysis) were treated with standard of care, which consists of amputation of the limb to remove the primary tumor followed by with adjuvant chemotherapy. Predictably, treatment generally was associated with increased survival, but the negative correlation with miR-134 and miR-154 was not driven by the type of treatment that the dogs received, and in fact, the data suggest that the dogs that received palliative care and amputation (2, 8 and 58 days survival) would have been reasonable candidates for more prolonged survival with standard of care treatment.

We have therefore identified and confirmed an association between metastasis and 14q32 miRNA expression levels in human osteosarcoma. Further, we have shown that the transcript reductions and prognostic significance for 14q32 miRNAs are conserved in canine osteosarcoma. Based on these integrative, comparative, multi-species analyses, we conclude that larger decreases in 14q32 miRNA expression levels are associated with an increased likelihood of metastases and poor outcomes in osteosarcoma patients.

Thus, this disclosure provides a method that generally includes analyzing a tissue sample from a subject for expression of a 14q32 miRNA. The method can be used as a diagnostic and/or prognostic tool. As a diagnostic tool, the method can inform one whether the subject from which the analyzed sample is obtained has or is at risk of having

osteosarcoma. Generally, if the tissue exhibits reduced expression of the 14q32 miRNA compared to a normal control, then the subject has or is at risk of having osteosarcoma.

As a prognostic tool, the method can be used to distinguish, to a clinically relevant extent, whether the osteosarcoma is an aggressive osteosarcoma— i.e., one that exhibits one or more aggressive behaviors such as, for example, metastasis, which can result in shorter survival times— or a nonaggressive osteosarcoma. Generally, if the expression of the 14q32 miRNA is no more than a predetermined percentage of 14q32 miRNA expression compared to a normal control, then the subject has or is at risk of having an aggressive osteosarcoma. With such a prognosis, an appropriate course of treatment can be implemented. Thus, for example, the method can further include treating a subject with treatment appropriate for an aggressive osteosarcoma if the subject exhibits expression of an 14q32 miRNA that is no more than the predetermined percentage of 14q32 miRNA expression indicative of being at risk for aggressive osteosarcoma. As another example, the method can further include treating a subject with treatment appropriate for a nonaggressive osteosarcoma if the subject exhibits expression of an 14q32 miRNA that is greater than the predetermined percentage of 14q32 miRNA expression indicative of an aggressive osteosarcoma and is less than the

predetermined 14q32 miRNA expression level of a normal control.

The standard of care aggressive and nonaggressive osteosarcoma are presently similar because practitioners are not currently able to distinguish between aggressive and

nonaggressive osteosarcomas. In humans, the standard of care typically includes 10 weeks of pre-operative chemotherapy followed by surgery that preserves the limb and 20 weeks of post-operative chemotherapy. Conventional chemotherapeutic agents including, for example, doxorubicin and cisplatin are currently being used in this treatment. These drugs, along with methotrexate, ifosfamide, or in combination with leucovorin, are often used to treat osteosarcoma in young adults. Percent necrosis in the tumor is used as a predictor for response, but the intensity of therapy is not adjusted based on this biomarker. Moreover, the predictive value of the tumor necrosis response can be as low as 50%. Consequently, practitioners do not have a sufficiently reliable indicator that a particular osteosarcoma is nonaggressive that may allow the practitioner to confidently moderate treatment. Instead, practitioners are reluctant to moderate treatment and will, instead, prescribe treatment that is consistent with the osteosarcoma being aggressive.

In dogs, the standard of practice is amputation of the limb, or in those cases where it is indicated, limb sparing surgery to remove the primary tumor followed by adjuvant chemotherapy. Pre-operative chemotherapy is not generally used in dogs. Post-operative chemotherapy generally is restricted to use of a single drug, with the most common agents used being doxorubicin and carboplatin. Recent studies also have investigated the use of these two drugs in combination with mixed results. There are no tests currently available to predict response to therapy in canine osteosarcoma, so all dogs receive the same intensity of chemotherapy irrespective of potential differences in tumor behavior.

As a prognostic tool, therefore, methods provided herein can allow one to better tune a patient's treatment for osteosarcoma based on the type of osteosarcoma affecting the patient. For example, the methods provided herein can allow a practitioner to distinguish between aggressive and nonaggressive osteosarcomas with greater confidence that is possible using existing methods. This can permit a practitioner to moderate treatment for patients with nonaggressive osteosarcomas. The moderated treatment may involve prescribing smaller doses of chemotherapeutic drug and/or less frequent chemotherapy, each of which can reduce undesirable side effects of the chemotherapy. Alternatively, the moderated treatment can involve the use of alternative chemotherapeutic agents that may be effective for reducing nonaggressive osteosarcomas but are considered ineffective for reducing aggressive osteosarcomas.

As used herein, the term "expression" and variations thereof refer to the transcription of a R A (e.g., a microRNA (imRNA) from DNA. Also as used herein, the term "at risk" refers to a subject that may or may not actually possess the described risk. Thus, for example, a subject "at risk" of a neoplastic condition is a subject that exhibits one or more risk factors associated with the condition— e.g., miRNA expression, genetic predisposition, ancestry, age, sex, geographical location, lifestyle, or medical history— even if the subject does not exhibit any clinical sign or symptom of the neoplastic condition.

In some cases, the decrease in 14q32 miRNA expression compared to a normal control can be expressed in terms of a predetermined percentage of 14q32 miRNA expression compared to the normal control. In certain embodiments, 14q32 miRNA expression that indicates a subject that has or is at risk of having osteosarcoma can include 14q32 miRNA expression that is a maximum of no more than 90% of a normal control such as, for example, no more than 80%, no more than 70%, no more than 60%, no more than 50%, no more than 40%, no more than 35%, no more than 30%, no more than 25%, no more than 20%, no more than 15%, no more than 14%, no more than 13%, no more than 12%, no more than 11%, no more than 10%, no more than 9%, no more than 8%, no more than 7%, no more than 6%, no more than 5%, no more than 4%, no more than 3%, or no more than 2% of a normal control. In certain embodiments, 14q32 miRNA expression that indicates a subject that has or is at risk of having osteosarcoma can include, while still exhibiting a decrease compared to a normal control, 14q32 miRNA expression that is a minimum of at least 1% of 14q32 miRNA expression in the normal control such as, for example, at least 2%, at least 3%, at least 4%, at least 5%, at least 6%, at least 7%, at least 8%, at least 9%, at least 10%, at least 11%, at least 12%, at least 13%, at least 14%, at least 15%, at least 16%, at least 17%, at least 18%, at least 19%, or at least 20% of the normal control. The 14q32 miRNA expression that indicates that the subject has or is at risk of having an osteosarcoma also can be characterized by any range that includes, as endpoints, any combination of a minimum 14q32 miRNA expression identified above and any maximum 14q32 miRNA expression identified that is greater than the minimum 14q32 miRNA expression.

In one embodiment, therefore, 14q32 miRNA expression indicative of a subject having or at risk of having an osteosarcoma can be, for example, no more than 40% of a normal control sample.

In another embodiment, 14q32 miRNA expression indicative of a subject having or at risk of having an aggressive osteosarcoma can be, for example, no more than 10% of a normal control sample.

As used herein, "aggressive osteosarcoma" refers to a form of osteosarcoma that exhibits one or more characteristics that are customarily associated with shorter survival times compared to other forms of osteosarcoma. Exemplary characteristics include, for example, metastasis, elevated expression of pro-survival and/or metastatic genes (e.g., TWIST 1, cMYC, and RUNX2), and/or decreased expression of tumor suppression genes (e.g., PTEN, FasR, ST5, ST7, ST14, and YPEL3).

As used herein, "nonaggressive osteosarcoma" refers to a form of osteosarcoma that does not exhibit aggressive behaviors (e.g., indolent).

In some embodiments, the 14q32 miRNA can include one or more of miR-382, miR- 134, miR-544 miR-379, miR-1197, miR-329-1, miR-494, miR-495, miR-376C, miR-1185-1, miR-1185-2, miR-381, miR-655, miR-487A, miR-496, miR-154, miR-369, miR-127, miR- 299, miR-665, miR-770, miR-432, miR-668, miR-539, miR-433, miR-323, miR-496, miR- 654, miR-409, miR-380, miR-485, miR-136, miR-376, miR-431, miR-370, miR-889, miR- 381, miR-299, miR-379, miR-411, miR-655, miR-432, miR-487, miR-410, miR-493, miR- 656, miR-485, miR-377 or miR-541.

Expression of the 14q32 miRNA may be determined using any suitable methods.

Conventional methods for measuring miRNA expression include, for example, qRT-PCR, microarrays, and RNA sequencing. In one particular embodiment, the miRNA expression is measured using qRT-PCR. In another particular embodiment, the miRNA expression is measured using a microarray. In another particular embodiment, the miRNA expression is measured using RNA sequencing. In the preceding description, particular embodiments may be described in isolation for clarity. Unless otherwise expressly specified that the features of a particular embodiment are incompatible with the features of another embodiment, certain embodiments can include a combination of compatible features described herein in connection with one or more embodiments.

For any method disclosed herein that includes discrete steps, the steps may be conducted in any feasible order. And, as appropriate, any combination of two or more steps may be conducted simultaneously.

The term "and/or" means one or all of the listed elements or a combination of any two or more of the listed elements; the terms "comprises" and variations thereof do not have a limiting meaning where these terms appear in the description and claims; unless otherwise specified, "a," "an," "the," and "at least one" are used interchangeably and mean one or more than one; and the recitations of numerical ranges by endpoints include all numbers subsumed within that range (e.g., 1 to 5 includes 1, 1.5, 2, 2.75, 3, 3.80, 4, 5, etc.).

The present invention is illustrated by the following examples. It is to be understood that the particular examples, materials, amounts, and procedures are to be interpreted broadly in accordance with the scope and spirit of the invention as set forth herein.

EXAMPLES

Human osteosarcoma tissue samples.

Human tissues were obtained from two sources. Frozen osteosarcoma tissue samples and normal bone samples (femur/ tibia) of individuals from similar age groups were obtained through the tissue procurement facility at the University of Minnesota (Bionet). Demographic and clinical information is provided in Tables 2 and 3. The excised primary osteosarcoma tumors were obtained prior to the initiation of chemotherapy or radiotherapy with informed consent and following institutional review board-approved protocols. Surgical resections were subjected to standard clinical and histopatho logical evaluation of hematoxylin and eosin (H&E) stained sections. An additional 16 human osteosarcoma samples with complete clinical follow-up information, including outcomes, were obtained from Pancras C.W.

Hogendoorn with approval from the institutional review board in compliance with the declaration of Helsinki. Demographic and clinical information is provided for these samples in Table 4.

Table 3. Demographic and clinical information of normal bone samples used in the study.

FT Age Race Gender Source

FT-401 20 Caucasian F Femur

FT-402 18 Caucasian F Femur

FT-403 15 Caucasian M Tibia

FT-404 6 Caucasian F Femur

FT-405 11 Caucasian F Tibia

FT-406 21 Caucasian M Femur

Table 2. Demographic and clinical information of human OS patient samples.

FT # Diagnosis Age Sex Race Tumor location Metastatic status Additional comments

FT-1 osteosarcoma 16 M white right tibia no conventional type, high grade

FT-2 osteosarcoma 16 F white Mediastinum (bronchus) Yes aortic dissenction, margins negative

FT-3 osteosarcoma 10 M hispanic right dital femur no conventional type, high grade

FT-4 osteosarcoma 64 M white soft tissue of neck no patient diseased due to metastasis in brain

FT-5 osteosarcoma 40 M white tibia no high grade OS in the proximal tibia

FT-6 osteosarcoma 17 M white tibia no conventional type, high grade

FT-7 osteosarcoma 16 M black femur no conventional type in distal femur

FT-8 osteosarcoma 12 F white left distal femur no conventional type, high grade

FT-9 osteosarcoma 19 M white left lung yes conventional type, high grade

FT-10 osteosarcoma 19 M white left lung yes conventional type, high grade

FT-11 osteosarcoma 25 F white right knee no soft tissue of the knee, recurrent

FT-12 osteosarcoma 57 M white right lung, lower lobe yes NA

FT-13 osteosarcoma 24 M white right chest wall yes NA

FT-14 osteosarcoma 14 M white left distal femur no NA

FT- 15 osteosarcoma 17 M white left distal femur no high grade OS

FT-16 osteosarcoma 22 M white left lung, lower lobe Yes metastatic OS extending to pleural surface

FT- 17 osteosarcoma 31 F white left lung, upper lobe yes vimentin+, actin+

extraskeletal OS with focal telangiectatic

FT- 18 osteosarcoma 42 M white right forearm no feautures

FT-19 osteosarcoma 31 F American Indian right lung, upper lobe yes osteoclasts present

Right pericardial

FT-336 osteosarcoma 20 M white mediastinal mass yes NA

FT-337 osteosarcoma 24 M white Left distal femur no NA

FT-338 osteosarcoma 20 M white Right chest wall yes NA

FT-339 osteosarcoma 24 M white Tumor, left femur no high grade, left femur

FT-340 osteosarcoma 20 M white Right middle lobe lung yes NA

Table 4. Human OS patient sample information and clinical follow-up.

Age Tumour Huvos pre-operative

PA-number Age* years Sex Subtype Location grade Metastasis ToM Deceased ToD ToF chemotherapy

H90-02350 200 16.7 M Telangiectatic distal femur right 1 yes 0 yes 25 25 PIA

H92- 10246 37 3.1 F Osteoblastic proximal tibia left 4 no no 194 PIA

H96-04177 205 17.1 F Osteoblastic distal femur left 1 no no 147 AP

H96-10281 304 25.3 M Osteoblastic proximal tibia left 2 yes 27 yes 47 47 AP proximal humerus

H98-13491 200 16.7 M Osteoblastic left 1 yes 0 yes 11 11 unknown

H00-03732 198 16.5 M Osteoblastic proximal tibia left 3 yes 10 yes 33 33 AP

HO 1-04259 244 20.3 F Anaplastic distal femur right 3 no no 94 AP

H02-00070 223 18.6 M Chondroblastic proximal tibia left 3 yes 21 yes 39 39 AP

H02-01830 137 1 1.4 F Osteoblastic proximal tibia left 2 no no 77 AP proximal tibia

H05-09073 181 15.1 M Osteoblastic right 3 no no 40 MAP

H06-06362 181 15.1 F Osteoblastic proximal tibia left 4 no no 32 MAP

H06-07381 96 8.0 M Osteoblastic distal femur left 2 yes 0 no 31 MAP

H07-01741 103 8.6 M Unknown distal femur right 3 yes 2 no 31 MAP

Osteoblastic,

H07-04382 164 13.7 M some giant cells distal femur right 3 no no 24 MAP

H07-08899 165 13.8 F Osteoblastic distal femur right 3 no no 24 MAP

Osteoblastic, proximal humerus

H08-02814 114 9.5 F some giant cells left 3 no no 17 MAP

*Age in months

ToM time of metastasis (months)

ToD time of death (months)

ToF time of follow up (months)

PIA Platinum, Ifosfamide, and Adriamycin

AP Adriamycin and Platinum

MAP Methotrexate, Adriamycin, and Platinum

Canine osteosarcoma tissue samples and cell lines.

The procedures to obtain canine osteosarcoma samples and establish corresponding explants and canine cell lines have been described previously (Scott et al, Bone 2011, 49:356-367; Thomas et al, Genome Res 2005, 15: 1831-1837). Briefly, na ^' ive (prior to treatment) samples were obtained from family-owned pet dogs with stage I or stage II appendicular osteosarcoma. Twenty-three of the 37 canine osteosarcoma samples used in this study were derived from the primary tumor in the leg. Samples OSCA-6 (Osteosarcoma Cell ACCR-6), OSCA-11, and OSCA-16 were derived from pulmonary metastases. Participation in the study required the dogs' owners to sign an informed consent form indicating they understood the goals and procedures for this study. Protocols and procedures were reviewed by the appropriate Institutional Animal Care and Use Committee. Samples were collected by attending veterinarians under sterile conditions as part of a diagnostic biopsy procedure, immediately after limb amputation, or at necropsy. Grossly visible tumor was dissected from adjacent normal tissue, rinsed in sterile phosphate buffered saline solution and divided into three portions that were (1) fixed in 10% neutral buffered formalin, (2) snap frozen, and (3) disaggregated into single cell suspensions. Histologically, the cellular composition of the resulting trimmed samples consisted of >90% tumor cells. For the single cell suspensions, debris was removed by passing sequentially through 500- and 200-μιη mesh filters. Cells were then placed in a 4-cm ² culture dish at a density of 1 x 10 ⁶ /ml in 1 ml of DMEM supplemented with 10% fetal bovine serum and cultured at 37°C in a 5% C0 ₂ atmosphere. Cells were allowed to reach ~80%> confluence, at which time they were passed to a 10-cm culture dish, and then again to a 25 -cm culture flask. By the second passage, the cultures were homogeneous and consisted of large polygonal to plump spindloyd or slightly rounded cells. Histologic diagnosis of osteosarcoma was verified for all tumors from formalin- fixed tissues. Osteoblastic origin of the cultured cells was confirmed at the third passage by expression of alkaline phosphatase and osteocalcin. As needed, cells were maintained in culture using the same conditions described above. Selected cell lines were tested after 30 passages in culture and had no noticeable changes in morphology or loss of alkaline phosphatase expression. These cell lines are tumorigenic in vivo, and, when injected orthotopically into nude mice, they have the capacity to metastasize spontaneously to the lungs (Wolfe et al., Clinical Exp Metastasis 2011, 28:377-389). Demographic and clinical information for the canine samples is provided in Table 5.

Table 5. Clinical information of canine OS cases.

Age at Treatment Survival Experi¬

Case ID Breed Gender Diagnosis Morphology Type (mo) ment

OSCA-1 Rottweiler F 9 Osteoblastic PALL 0.99 GEP

OSCA-04 Rottweiler F 9.91 Osteoblastic PALL 0.26 Q

OSCA-6 Rhodesian Ridgeback F 2.4 Osteoblastic PALL 0.36 GEP

OSCA-8 Rottweiler M 2 Osteoblastic NA LTF GEP

OSCA-11 Rottweiler M 5 Chondroblastic NA LTF GEP

OSCA-12 Rottweiler MC 7 Osteoblastic NA 1 GEP

OSCA-16 Rottweiler M 6.3 Osteoblastic soc 1.97 GEP

OSCA-17 Rottweiler M 8.9 Fibroblastic soc 64.01 GEP

OSCA-19 Rottweiler MC 8 Chondroblastic soc 10.52 GEP

OSCA-20 Golden Retriever MC 7 Osteoblastic soc 1.71 GEP

OSCA-21 Rottweiler MC 11.2 Fibroblastic soc 13.12 GEP

OSCA-22 Rottweiler MC 9.1 Osteoblastic AMP 0.23 GEP

OSCA-23 Golden Retriever MC 8.9 Osteoblastic SOC 13.97 GEP

OSCA-25 Rottweiler FS 7.1 Fibroblastic SOC 4.02 GEP

OSCA-29 Rottweiler F 6.8 Osteoblastic soc 2.33 GEP/Q

OSCA-30 German Shepherd MC 8 Osteoblastic soc 20.61 GEP/Q

OSCA-32 Great Pyrenees F 8.7 Fibroblastic soc 11.01 GEP

OSCA-33 Rottweiler FS 12.08 Osteoblastic soc 13.55 Q

OSCA-40 Saint Bernard FS 5.8 Osteoblastic soc 1.18 GEP

Osteomyelitis

OSCA-41 * Alaskan Malamute M (no tumor) AMP >36 GEP/Q

OSCA-45 Irish Setter FS 11 Osteoblastic AMP 5.16 Q

OSCA-50 Rottweiler M 3.7 Osteoblastic PALL 0 GEP

OSCA-55 Borzoi F 11.91 Osteoblastic NA 1.87 Q

OSCA-56 Rottweiler M Unknown Osteoblastic AMP LTF GEP

OSCA-57 Golden Retriever M 7.8 Osteoblastic SOC 7.3 GEP/Q

OSCA-59 Golden Retriever FS 9.4 Chondroblastic SOC 38.96 GEP

OSCA-60 Mastiff F 6 Osteoblastic SOC 4.96 GEP

OSCA-67 Mastiff FS 12.27 Mixed PALL 0.07 Q

OSCA-71** Golden Retriever FS 6.7 Osteoblastic SOC 16 GEP

OSCA-73 Golden Retriever FS 2.5 Osteoblastic PALL 1.48 GEP

OSCA-75 Golden Retriever FS 9.8 Osteoblastic SOC 7.66 GEP/Q

OSCA-77 Belgian MC 6.1 Osteoblastic SOC 4 Q

Mixed

Osteoblastic-

OSCA-78 German Shepherd M 9.5 fibroblastic SOC 2.5 GEP/Q

OSCA- 84 Rottweiler FS 10 NA SOC 6.48 Q

OSCA- 96 Mastiff FS 5.17 NA SOC 11 Q OSCA- 98 Labrador Retriever MC 11 Osteoblastic soc 4.38 Q

OSCA- 99 Labrador Retriever Mix FS 10 Osteoblastic soc 4.35 Q

OSCA- 109 Irish Setter M 10.08 Osteoblastic soc 1.58 Q

OSCA = Osteosarcoma Cell line Animal Cancer Care and Research Program

NA = information not available in the record

LTF = lost to follow-up

*Final diagnosis was osteomyelitis

"Dog was alive at end of study

# canine OS samples used for qRT-PCR and survival analysis are highlighted

RNA isolation.

Total RNA was isolated from 75-100 mg of frozen human osteosarcoma tissue using the miRvana total RNA isolation kit (Ambion Inc, Austin TX, USA) following the manufacturer's protocol. RNA was quantified using the Nanodrop 8000 (NanoDrop

Technologies LLC, Wilmington, DE, USA). Samples with RNA Index Number (RIN) values of >6 were included in this study. Total RNA was isolated from canine osteosarcoma tissue using TriPure Isolation Reagent (Roche, Mannheim, Germany). The concentration of RNA from canine osteosarcoma tissue was determined using NanoDrop 1000 UV-Vis

spectrophotometer (NanoDrop Technologies LLC, Wilmington, DE, USA) and quality was assessed on a 1.2% formaldehyde agarose gel with ethidium bromide staining. RNA from canine osteosarcoma cryopreserved cells was isolated, quantified, and assessed for quality as previously described (Scott et al, Bone 2011, 49:356-367).

Analysis of osteosarcoma mRNA and miRNA expression.

Messenger RNA expression profiles from human tumors were generated using the human HT-12 Beadchip (Illumina Inc., San Diego, CA, USA) (Fan et al., Genome Res 2004, 14:878-885). Human miRNA expression profile datasets described previously (S-MED; Sarver et al, Laboratory Invest 2010, 90:753-761) were used in this study. Additional clinical information relating to these samples is described in Table 2 (human). Canine osteosarcoma mRNA gene expression data generated on Affymetrix canine_2.0 arrays were also used as previously described (Scott et al, Bone 2011, 49:356-367). Complete clinical and breed information for canine osteosarcoma data is provided in Table 5. miRNAs from the 14q32 cluster were mapped to the canine genome. Two miRNAs (miR-134 and miR-544) that showed 100% conservation and mapped to the predicted region of synteny in canine chromosome (CFA) 8 were used to examine expression of the 14q32 cluster in dog samples (FIG. 5).

Table 6. Differentially expressed mRNA in osteosarcoma positively correlated to 14q32 encoded miR-382.

Correlation to MiR-382 Correlation to MiR-382

Gene Name Gene Name

expression level expression level

MRGPRF 0.96 LOC653598 0.773

RGMA 0.928 CLEC1A 0.773

RGS11 0.928 DKFZP434K191 0.773

TNFRSF25 0.925 KLHL3 0.772

MEG3 0.924 LOC196752 0.772

LIMS2 0.924 RARRES2 0.771

HOXA6 0.917 SPON1 0.77

TAGLN 0.909 COBLL1 0.77

CXCL12 0.901 LOC644634 0.768

ADCY4 0.894 SYT11 0.768

LRRC4C 0.894 PRICKLE2 0.767

RASL12 0.892 OR10T2 0.766

NR2F2 0.89 COL8A1 0.766

FLJ35258 0.889 SETBP1 0.765

MEOX1 0.887 MYOZ3 0.765

PRICKLE 1 0.886 PPP1R13B 0.765

CLEC3B 0.885 LOC646897 0.765

GJC2 0.883 RAXLl 0.764

IL1RAPL1 0.877 NET1 0.764

LRRC32 0.876 RNF165 0.764

PHLDB1 0.873 CNN1 0.764

TFIP11 0.87 SOX 17 0.763

RAMP3 0.868 TMEM30B 0.763

DARC 0.867 PPP2R3A 0.762

LOC647654 0.867 LOC390876 0.761

FLJ45337 0.866 LOC390414 0.76

COX6BP1 0.866 KLF2 0.76

PEG3 0.866 GVIN1 0.76

TEK 0.865 IGFBP5 0.76

LOC649088 0.864 OTX1 0.76

LOC647873 0.862 MYOM1 0.759

COMP 0.86 LOC643272 0.759

ZNF385D 0.86 NOSTRIN 0.759

RPH3A 0.858 KIT 0.758

LOC643700 0.855 DLG4 0.758

LOC727948 0.855 MALL 0.758

FAM124B 0.854 PODN 0.757

LOC642321 0.853 S100A16 0.756

LOC643423 0.852 PTPRK 0.755

CD79B 0.851 FOXD4L1 0.755

DKFZP686K1684 0.85 PCDHB9 0.753 LOC645241 0.85 LOC441773 0.752

LOC651686 0.849 FLJ21438 0.751

NOL10 0.849 LOC643453 0.75

DEFB109 0.849 PROM1 0.749

LOC649181 0.849 LOC653524 0.749

NCALD 0.848 LOC728411 0.748

IGFL3 0.847 LOC199800 0.748

LOCI 52024 0.846 SULF1 0.748

LOC643605 0.845 ZNF521 0.747

MYH11 0.844 NISCH 0.747

CPXM1 0.844 PPA2 0.746

CCDC149 0.844 CCDC140 0.746

PTPRB 0.843 LOC642707 0.746

LOC651821 0.843 GPR116 0.745

LOC642118 0.841 ALS2CR14 0.744

IFNB1 0.84 SELP 0.743

KCNMB1 0.839 ESAM 0.742

SGIP1 0.838 LOC644642 0.742

FOXOl 0.838 ZNF160 0.742

FGF9 0.837 LOC644334 0.74

LOC648879 0.837 ARID5B 0.739

PDLIM3 0.836 IGFBP7 0.739

TSHZ2 0.835 DCLRE1C 0.737

LPPR4 0.834 AHI1 0.737

RDH5 0.834 RGS16 0.736

LOC647786 0.832 SLC4A5 0.735

A4GALT 0.831 ABLIM1 0.735

LOC646491 0.828 AHR 0.734

CPT1C 0.827 C80RF37 0.733

GRK5 0.826 C10ORF10 0.732

LOC642017 0.826 ABI3BP 0.732

LOC649456 0.826 ANKRD30B 0.731

STMN2 0.825 COX19 0.731

KRTAP21-2 0.824 C210RF55 0.731

CFB 0.824 EVI5 0.731

CARD 10 0.823 AGTRL1 0.731

NFIB 0.823 EVI1 0.73

ADCY2 0.821 PDSS2 0.73

FEZ1 0.818 CLEC14A 0.73

DLC1 0.815 SOX18 0.73

SLC02A1 0.815 TGFBR3 0.73

PGM5 0.815 LOC651268 0.729

NLGN2 0.815 SDHALP1 0.729

SPARCL1 0.814 APCDD1L 0.729

C9ORF106 0.814 KCNJ8 0.729

CACNA1H 0.814 CI S 0.728

RAPGEF3 0.814 PDCD4 0.728

GJA4 0.812 PRSS23 0.728

MGC61598 0.812 CAMK2N1 0.727

DLL1 0.811 COL15A1 0.727

LOC390033 0.81 TM4SF1 0.727 C10ORF108 0.809 LOC648059 0.727

PPP1R14A 0.809 LOC441019 0.726

LDB2 0.809 FBLN2 0.726

C20ORF46 0.809 ClORFl lO 0.725

LOC649661 0.806 C140RF78 0.725

PIK3C2B 0.806 THBS4 0.725

LOC645362 0.805 CPLX1 0.725

COL18A1 0.804 PDZRN4 0.724

TESC 0.803 LOC653489 0.724

FOXD4 0.803 LOC440157 0.723

FGD5 0.803 C10ORF116 0.722

TTN 0.802 GIMAP5 0.72

GABRB1 0.802 LOC653829 0.719

PALM 0.802 RASIP1 0.719

CD34 0.801 RUNDC2C 0.719

LOC648907 0.801 C10RF115 0.719

MGC10997 0.801 LOC389517 0.719

GAS 6 0.801 CLDN5 0.718

CAV1 0.799 ITPR3 0.718

NUAK1 0.798 PNPLA7 0.717

LOC643466 0.798 ZNF223 0.717

LOC643817 0.797 FAM153B 0.717

NOX4 0.797 CRIP2 0.716

LOC647295 0.796 ETS2 0.714

ACACB 0.795 ICA1 0.714

ACTG2 0.794 CEP27 0.714

LOC651309 0.794 TMEM16A 0.713

PPP2R2B 0.793 C8A 0.712

SPRYD5 0.792 PECAM1 0.712

DNHL1 0.792 ACTN4 0.711

ANGPTL2 0.791 VCL 0.711

MFAP5 0.791 FM03 0.71

CLDN14 0.791 SALL3 0.708

SMOC2 0.791 ARHGAP22 0.708

LOC389634 0.791 EIF2AK4 0.707

LOC653800 0.79 LOC653086 0.707

JAG1 0.79 PLA2G2D 0.707

KIAA0355 0.788 LOC728014 0.706

LOC644113 0.788 SH3BGRL2 0.706

SHANK3 0.786 RAB3IP 0.705

C60RF189 0.785 CDAN1 0.704

LOC653629 0.785 TIE1 0.703

C110RF38 0.784 NUBPL 0.703

TLE2 0.784 LOC123688 0.702

LOC645489 0.784 C9ORF80 0.702

MGC24103 0.782

SEMA6A 0.782

AKR1C3 0.781

EPHA4 0.781

LOC731895 0.779

ARHGEF3 0.778 LOC143666 0.777

FLJ41423 0.776

THY1 0.776

LOC642299 0.775

CASQ2 0.774

LOC641989 0.774

RHOJ 0.774

C20ORF160 0.773

For human samples, mRNA data were assayed for quality control as described (Sarver et al., J Cardiovasc Transl Res 2010, 3:204-211). Fluorescence values were obtained from the Illumina detection system without background subtraction and were quantile normalized using GeneData Expressionist Software (GeneData Inc, San Francisco, CA, USA). Multiple probes to the same gene product were then averaged to obtain a value for each gene. The dataset was filtered to identify genes with a variance >1 across the human osteosarcoma tumor dataset and two normal bone samples. The mRNA expression data were composed of human tumors and normal bone samples which had been previously profiled for miRNA expression levels, allowing us to directly calculate Pearson correlation coefficients between miR-382 and mRNA transcript levels. mRNAs that showed positive correlation >0.7 are listed in Table 6, and mRNAs that showed negative correlation are listed < -0.7 in Table 7. These mRNAs and their direction of change (increased for negative correlation and decreased for positive correlation) were submitted to Ingenuity Pathway Analyses to determine enriched functions where the direction of change was consistent with increased or decreased activity of enriched functions. The quantile normalized human miRNA array data were first filtered to determine miRNAs that showed variance >0.1 across the 15 human osteosarcoma tumors. Pearson correlations were then calculated between all pairs of miRNA. All pairs of miRNA with R >0.9 were then used to generate a correlation network visualization of the

osteosarcoma miRNAome. Pairwise Pearson correlations used to generate FIG. 1 A are provided in Table 8.

Table 7. Differentially expressed mRNA in osteosarcoma negatively correlated to 14q32 encoded miR-382.

Correlation to MiR-382 „ _{- T} Correlation to MiR-382 rene Name . . . rene Name . . .

expression level expression level

C140RF142 -0.892 UBE2C -0.741

PRKCA -0.884 EXOSC4 -0.741 PTRH2 -0.88 C9ORF105 -0.74

PIGP -0.873 STRAP -0.738

CALU -0.87 LOC134997 -0.737

POLR2J3 -0.863 G3BP2 -0.736

GPC4 -0.861 CLDN12 -0.733

C70RF49 -0.86 HIRIP3 -0.733

CCDC86 -0.849 KDELR2 -0.733

P4HA2 -0.844 BYSL -0.73

PSMD14 -0.84 TWIST 1 -0.729

MRPS18C -0.839 HMGN2 -0.729

MRPS33 -0.835 IARS -0.729

LASS2 -0.827 TYMS -0.725

TRIB3 -0.823 C130RF34 -0.725

ATP1B1 -0.816 SDHB -0.725

SPR -0.808 PFN2 -0.724

NEU1 -0.803 METTL1 -0.722

B4GALNT1 -0.803 MRPL40 -0.721

SDC2 -0.799 FEN1 -0.719

PPIH -0.799 RPN2 -0.719

RAD51AP1 -0.795 GGH -0.717

NOP5/NOP58 -0.794 CMTM6 -0.717

POLQ -0.794 C19ORF10 -0.717

COQ2 -0.792 UBL5 -0.716

SDC4 -0.791 AARS2 -0.716

MRC2 -0.79 CKS2 -0.715

NOMOl -0.786 PCOLCE2 -0.714

PIP4K2C -0.782 EBP -0.714

HIST1H1C -0.781 C30RF58 -0.714

COMMD9 -0.778 UBE2T -0.713

MRPL16 -0.777 C140RF2 -0.711

PRDX4 -0.776 LOC654121 -0.709

UQCRQ -0.773 DCXR -0.709

PRICKLE4 -0.77 GCUD2 -0.708

HNRPC -0.767 CDK4 -0.707

C60RF125 -0.766 STX8 -0.707

MARS -0.766 M160 -0.706

TFRC -0.766 C20RF7 -0.705

EPDR1 -0.765 LOC646197 -0.702

ALDOA -0.763 CCDC99 -0.701

RUNX2 -0.762 HIST2H2BE -0.701

DCI -0.76 CDCA5 -0.701

PSMA3 -0.757

MRPL13 -0.757 TK1 -0.753

NDUFA1 -0.753

PET112L -0.751

P4HA1 -0.75

HIST1H2BK -0.749

PCNA -0.749

SNRPC -0.747

CDK5 -0.747

FAM96A -0.742

Table 8. 14q32 subnetwork pairwise correlations.

Row position

HS_212

HS_214.1 100584039

hsa-miR-

127 100419125

hsa-miR-

134 100590784

hsa-miR-

154 100595859

hsa-miR-

154* 100595895

hsa-miR-

299-3p 100559922

hsa-miR-

299-5p 100559890

hsa-miR-

323 100561871

hsa-miR-

329 100562924

hsa-miR-

368 100575823

hsa-miR-

369-3p 100601731

hsa-miR-

370 100447276

hsa-miR-

376a 100576208

hsa-miR-

376a* 100576877

hsa-miR-

376b 100576587

hsa-miR-

377 100598184

hsa-miR-

379

hsa-miR-

381 100582058

hsa-miR-

382 100590406

hsa-miR-

409-3p 100601435 hsa-miR-

410 100602051 hsa-miR-

411 100559430 hsa-miR-

431 100417116 hsa-miR-

432 100420586 hsa-miR-

485-3p 100591554 hsa-miR-

487a 100588584 hsa-miR-

487b 100582595 hsa-miR-

493-3p 100405206 hsa-miR-

493-5p 100405165 hsa-miR-

495 100569894 hsa-miR-

496 hsa-miR-

654 100576324 hsa-miR-

655 100585700 hsa-miR-

656 100602856

Integrative analysis of human and canine gene and miRNA expression data.

Human genes that showed positive and negative correlation to miR-382 were mapped to canine genes using gene symbols. The direction of change was tracked using dark grey (negative correlation) and light grey (positive correlation) tags for each gene so that the direction of change in the human data could be observed in analogous fashion as previously described (Scott et al, Bone 2011, 49:356-367). Unsupervised hierarchical clustering was carried out with the Pearson correlation as the metric using average linkage following log base 2 transformation of mean-centered data using Cluster3.0. Heatmaps were generated from the CDT files generated by Cluster3.0 using Java Treeview (Saldanha, Bioinformatics 2004, 20:3246-3248). qRT-PCR

qRT-PCR of RNA from both human and canine osteosarcoma samples was carried out using normalization to U6 snRNA. First strand cDNA was synthesized from total RNA using a miScript reverse transcription kit (Qiagen, Valencia, CA, USA). miRNA was quantified with an miRscript SYBR Green PCR kit (Qiagen, Valencia, CA, USA) using cDNA equivalent of 50 ng total RNA per reaction. Mature miRNA- specific forward primers provided in Table 1 were purchased from Integrated DNA Technologies (Coralville, IA, USA) and the universal reverse primer provided by the manufacturer.

Table 1. Oligonucleotides used for qRT-PCR analysis.

Real-time PCR was performed at 55°C annealing following standard protocol of the manufacturer in 7500 Real Time PCR system (Applied Biosystems/Life Technologies, Inc., Carlsbad, CA USA) and threshold cycles (CT) were calculated using Sequence Detection Software (SDS vl .2.1, Applied Biosystems/Life Technologies, Inc., Carlsbad, CA USA). Relative gene expression (expressed as fold difference relative to normal bone tissues (human OS) or relative to reactive osteoblasts (canine OS)) was calculated from the average of triplicate CT value measurements using the 2-AACt method (Schmittgen et al., Nat Protoc 2008, 3: 1101-1108) except for situations where the RNA available was limited allowing only duplicate reactions to be carried out. The standard error is represented as 2 ^_( - ^{ΔΔετ ± SO}

Statistical analyses

Kaplan-Meier survival curves were generated with the GeneData Analyst statistical package using the clinical data provided in Tables 4 and 5. Whitney Mann U tests were conducted using an online tool available at vassarstats.net/utest.html.

The complete disclosure of all patents, patent applications, and publications, and electronically available material (including, for instance, nucleotide sequence submissions in, e.g., GenBank and RefSeq, and amino acid sequence submissions in, e.g., SwissProt, PIR, PRF, PDB, and translations from annotated coding regions in GenBank and RefSeq) cited herein are incorporated by reference in their entirety. In the event that any inconsistency exists between the disclosure of the present application and the disclosure(s) of any document incorporated herein by reference, the disclosure of the present application shall govern. The foregoing detailed description and examples have been given for clarity of understanding only. No unnecessary limitations are to be understood therefrom. The invention is not limited to the exact details shown and described, for variations obvious to one skilled in the art will be included within the invention defined by the claims.

Unless otherwise indicated, all numbers expressing quantities of components, molecular weights, and so forth used in the specification and claims are to be understood as being modified in all instances by the term "about." Accordingly, unless otherwise indicated to the contrary, the numerical parameters set forth in the specification and claims are approximations that may vary depending upon the desired properties sought to be obtained by the present invention. At the very least, and not as an attempt to limit the doctrine of equivalents to the scope of the claims, each numerical parameter should at least be construed in light of the number of reported significant digits and by applying ordinary rounding techniques . Notwithstanding that the numerical ranges and parameters setting forth the broad scope of the invention are approximations, the numerical values set forth in the specific examples are reported as precisely as possible. All numerical values, however, inherently contain a range necessarily resulting from the standard deviation found in their respective testing measurements.

All headings are for the convenience of the reader and should not be used to limit the meaning of the text that follows the heading, unless so specified.