Soft tissue sarcomas (STS) are a heterogeneous group of mesenchymal tissue cancers, with over 50 histological sub-types. Regardless of type, virtually all STS are treated the same; i.e. with doxorubicin (DOX) followed by resection, if possible. More than 20% of STS are non-resectable and, of those that are, more than 20% recur. For non-resectable or recurring patients, median survival is a dismal 18 months. Perhaps because of its rarity, no new front-line agents have been developed for STS in decades. Among newer agents being developed for STS, TH-302 is showing exceptional promise. TH-302 is an alkylating pro-drug that is activated only in regions of severe hypoxia, and is currently in a phas III trial in combination with DOX in unresectable STS. The primary endpoint of this trial is overal survival with a secondary endpoint of objective radiological response with RECIST 1.1. RECIST responses rarely correlate with survival responses in soft tissue sarcomas, however. The purpose of this work is to develop MR imaging biomarkers that can predict response to DOX and/or TH-302, with the overarching hypothesis that distinct MRI-defined sub-regions of tumors will have differential responses to these agents. Because these sub-regions are defined by their distinct physiology illuminated by combining multiple MR scans, we have termed these as distinct "habitats". Because TH-302 is active only in hypoxia, and hypoxia should be represented by a specific habitat, we specifically hypothesize that imaging of the hypoxic habitat can be used to predict and monitor responses to TH-302. The hypoxia habitat is classified as having low perfusion and high cell density. In contrast, the DOX-responsive habitat should be well perfused (yielding higher drug concentrations) and have high cell density (with more drug targets). This represents a conceptual advance as such habitats may provide a common predictive biomarker across the multiple histological sub-types of STS for patient stratification and therapeutic decision support. The approach will follow on preliminary work, wherein delineated habitats were quantitatively identified across multiple histological types and grades ofSTS by combining T1, contrast- enhanced T1, and T2 STIR MR images. The current proposal is entirely pre-clinical, with the expectation that findings herein can be rapidly translated to clinial care. Preclinical work is justified in that there is greater flexibility to interrogate a wider porfolio of MRI pulse sequences and treatment strategies that can be related to underlying histology and molecular profiling. Aim 1 will quantitatively compare MR-visible habitats to histology and molecular profiles of xenotransplanted tumors. Aim 2 will test the hypothesis that tumors with different habitat profiles will be differentially responsive to DOX and/or TH-302. In Aim 3, we wil investigate the effects of metabolic perturbations to affect the hypoxic habitat and thus improve response to TH-302. At the end of this study, we will have developed a new set of MR imaging biomarkers for predicting and monitoring response in this heterogeneous group of diseases, with the expectation that these finding will inform a follow-on study in the clinic.