MicroRNA-mRNA networks define translatable molecular outcome phenotypes in osteosarcoma

Christopher B. Lietz, Cassandra C. Garbutt, William T. Barry, Vikram Deshpande, Yen‐Lin Chen, Santiago A. Lozano‐Calderón, Yao‐Yu Wang, Brian Lawney, David H. Ebb, Gregory M. Coté, Zhenfeng Duan, Francis J. Hornicek, Edwin Choy, G. Petur Nielsen, Benjamin Haibe‐Kains, John Quackenbush, Dimitrios SpentzosScientific 2020 · DOI

Abstract

Osteosarcoma remains a clinical challenge with stagnant treatment protocols for over two decades. This study presents the largest single-center translational analysis of osteosarcoma to date, utilizing microRNA sequencing of 74 biopsy samples alongside multi-omic data from a national cooperative group. We identified and validated prognostic microRNA (miRNA) signatures that define distinct molecular phenotypes independent of traditional clinicopathologic features. Integration with transcriptomic and epigenomic data, followed by machine learning-based pharmacogenomic analysis, revealed novel therapeutic candidates. These findings offer a path toward personalized treatment intensification or alternative regimens based on upfront molecular risk stratification.

Introduction

Osteosarcoma is the most common primary bone malignancy in children and young adults. Despite the initial success of MAP chemotherapy (methotrexate, adriamycin, and cisplatin) decades ago, survival rates for patients with relapsed or metastatic disease have not improved significantly. Currently, prognosis is primarily assessed via pathologic necrosis—a late-stage, semi-quantitative measure of chemotherapy response. There is a critical need for early, molecularly-driven biomarkers to identify high-risk patients at diagnosis to guide more intensive or alternative therapies.

Methods and Datasets

The study employed a multi-omic approach to move beyond single-gene markers, utilizing two primary cohorts:

  • MGH Cohort: 74 frozen biopsy samples from Massachusetts General Hospital, analyzed via small RNA sequencing.
  • NCI TARGET Cohort: 95 samples with multi-omic data (miRNA, mRNA-seq, and DNA methylation) used for validation.

Researchers mapped miRNAs to their experimentally verified mRNA targets and analyzed DNA methylation patterns to ensure network consistency. To ensure clinical translatability and avoid overfitting, they used a "Signed Averaged Expression" scoring method to create robust prognostic signatures. Finally, pharmacogenomic modeling via the PharmacoDB platform linked these signatures to drug sensitivity data across hundreds of cell lines to identify novel therapeutic candidates.

Results: Molecular Stratification of Survival

The analysis identified several miRNA signatures, many clustered on the 14q32 chromosomal locus, that significantly predict patient outcomes. These molecular phenotypes proved to be more predictive than traditional clinical features like metastatic status at diagnosis.

Kaplan-Meier survival analysis for recurrence-free survival (RFS) across different miRNA-defined risk groups (Panels B, D, F).

As shown in the survival curves, the miRNA profiles effectively stratify patients into distinct risk groups. In the MGH cohort, a 22-miRNA profile demonstrated a median RFS of 33 months for the high-risk group versus "not reached" for the low-risk group (p=0.032). Multivariate analysis confirmed the independence of these markers, with the 5-miRNA profile maintaining a Hazard Ratio of 6.1 even when adjusted for chemotherapy response and metastatic status.

Validation of prognostic miRNA signatures in independent cohorts showing consistent survival outcomes (Panels A, C).

The consistency of these results across different cohorts and platforms underscores the biological relevance of these networks. Furthermore, the integration with mRNA data showed that the targets of these miRNAs were also prognostic, suggesting a functional regulatory network driving tumor aggressiveness. The best predictive power was achieved by combining miRNA profiles with traditional pathologic necrosis, allowing for the stratification of patients into "very favorable," "intermediate," and "very unfavorable" groups.

Pharmacogenomic Insights and Future Directions

By linking molecular signatures to drug sensitivity data, the study identified 19 candidate drugs for repurposing. Beyond validating standard agents like doxorubicin, the model pointed toward CDK, MEK, and PI3K inhibitors as potential therapies for high-risk molecular subtypes.

While the study is retrospective, the robustness of the 14q32-associated signatures provides a strong foundation for prospective clinical trials. The ultimate goal is to use these molecular "compasses" at diagnosis to tailor therapy—sparing low-risk patients unnecessary toxicity while directing high-risk patients toward novel, targeted interventions like dasatinib or alvocidib.

Summary updated

Abstract

There is a lack of well validated prognostic biomarkers in osteosarcoma, a rare, recalcitrant disease for which treatment standards have not changed in over 20 years. We performed microRNA sequencing in 74 frozen osteosarcoma biopsy samples, constituting the largest single center translationally analyzed osteosarcoma cohort to date, and we separately analyzed a multi-omic dataset from a large NCI supported national cooperative group cohort. We validated the prognostic value of candidate microRNA signatures and contextualized them in relevant transcriptomic and epigenomic networks. Our results reveal the existence of molecularly defined phenotypes associated with outcome independent of clinicopathologic features. Through machine learning based integrative pharmacogenomic analysis, the microRNA biomarkers identify novel therapeutics for stratified application in osteosarcoma. The previously unrecognized osteosarcoma subtypes with distinct clinical courses and response to therapy could be translatable for discerning patients appropriate for more intensified, less intensified, or alternate therapeutic regimens.
BibTeX
@inproceedings{lietz2020microrna,
  author = {Christopher B. Lietz and Cassandra C. Garbutt and William T. Barry and Vikram Deshpande and Yen‐Lin Chen and Santiago A. Lozano‐Calderón and Yao‐Yu Wang and Brian Lawney and David H. Ebb and Gregory M. Coté and Zhenfeng Duan and Francis J. Hornicek and Edwin Choy and G. Petur Nielsen and Benjamin Haibe‐Kains and John Quackenbush and Dimitrios Spentzos},
  title = {{MicroRNA-mRNA networks define translatable molecular outcome phenotypes in osteosarcoma}},
  booktitle = {Scientific},
  year = {2020},
  doi = {10.1038/s41598-020-61236-3},
}
Figure