Leveraging Patient-Derived Xenograft Models to Inform Clinical Characterization of Sarcoma That Would Impact Treatment

Abstract

Sarcoma describes a rare family of cancers that arise in connective tissue which include fat, muscle, blood vessels, deep skin tissues, nerves, bone and cartilage. Sarcomas are typically described by two main types: soft tissue and bone. Soft tissue sarcomas comprise of less than 1% of all cancer diagnoses, and the American Cancer Society states that there will be 13,040 new cases of soft tissue sarcoma diagnoses. Within soft tissue sarcoma, there are over 50 described subtypes, but precise diagnosis is difficult due to lack of understanding of the biology and genetic background of these subtypes and the lack of resources for diagnostic tests. Regardless of diagnosis, though, there are few therapeutic options for these patients; better understanding of this disease and its subtypes could have the potential to identify new treatments or modify existing treatments for this rare and deadly disease. Due to the rarity and diverse nature of this cancer type, conducting large genomic studies to determine underlying genetic mechanisms to this disease have been challenging. This study sought to first evaluate the current diagnosis process by IHC using mRNA expression data from an RNAseq experiment of 78 sarcoma patient-derived xenograft (PDX) models. First we determined if gene expression can be used as a proxy to protein expression and second we tested if there were genes that are not evaluable by IHC can differentiate the subtypes better or define new subtypes that have more clinical relevance. We evaluated these models as to how well they represent human disease by comparing subtype composition and gene expression patterns to the Cancer Genome Atlas’s (TCGA) sarcoma study that comprised of 256 sarcoma patients. We then used supervised and unsupervised methods to identify sarcoma groupings that may help generalize the disease better than the subtypes, with a focus on identifying clinically relevant differential gene expression that can also have a direct impact on treatment options for sarcoma patients. The findings from these clustering experiments were also compared to TCGA samples to determine clinical relevance, or translatability of these models, and patient population for trial recruitment or stratification for sarcoma patients. In addition to confirming that RNA measurements can be used as a potential replacement for IHC, we discovered a subsets of sarcoma PDX models with increased expression of VEGF-C, PRAME, CTLA-4 and MAGE-A gene family members. While the corroborating evidence in TCGA is weak as these genes were only up-regulated in <10% of sarcoma patients, but the subtypes that showed mRNA up-regulation were of the same subtypes that were identified in PDX models. These findings do have potential to have therapeutic impact by exploring the potential use of pazopanib for the VEGFC over-expressing sarcomas or the CTLA-4 expressing subtype could benefit from immune therapies such as ipilimumab. In general, this study showed the value of preclinical models and their ability to reflect the human disease. Future work in this area of sarcoma should include correlating response to targeted and immune-oncology treatment with genes of interest in sarcoma patients to inform on patient selection and stratification strategies.