Immune and muli-omic profiling for molecular classification and biomarker discovery in soft tissue sarcoma

Abstract

Soft tissue sarcomas (STS) are a group of over 50 rare cancers of mesenchymal origin. Classification of STS is primarily based on histological characteristics of tumours. Novel biomarkers are required that account for the heterogeneity in clinical phenotype and underlying cancer biology seen within and between histological STS subtypes. To address the unmet need for new biomarkers, the studies reported in this thesis assembled clinically annotated tumour cohorts for use in tissue profiling experiments. First, we aimed to characterise the immune tumour microenvironment (iTME) of leiomyosarcoma (LMS), undifferentiated pleomorphic sarcoma (UPS) and dedifferentiated liposarcoma (DDLPS) - three STS subtypes characterised by karyotypic complexitiy - and to assess for association between immune phenotype and clinical outcome. We performed TMA-based IHC profiling of infiltrating immune cells and expression analysis of 21 immune-related genes in a retrospective series of 266 early stage tumours across these 3 subtypes. We found substantial quantitative variation in immune cell infiltration between individual tumours, although differences between the 3 subtype-defined subcohorts were not pronounced. Significant association between improved overall survival and denser immune cell infiltrates was seen in UPS, while no such associations were seen in LMS and DDLPS. Meanwhile, the combination of unsupervised clustering of immune gene expression data with associated IHC-based data, led to the identification of 4 distinct immune-based subgroups that exhibited qualitatively and quantatively contrasting iTME characteristics, were not restricted by conventional histological subtype classification and were associated with significant differences in prognosis. These hypothesis-generating results indicate that iTME-based prognostic biomarkers warrant further development as risk classifiers in early stage disease both within and across subtypes of karyotypically-complex STS. The clinical effectiveness of pazopanib, a multitargeted kinase inhibitor with anti-angiogenic and anti-oncogenic activity, in the treatment of advanced STS is limited by the lack of predictive biomarkers and a poor understanding of the clinical mechanisms of drug response and resistance. In a retrospective series of pazopanib-treated patients with advanced STS, we undertook molecular profiling of pre-treatment tumour specimens which led to the development of an integrated PARSARC (Pazopanib Activity and Response in SARComa) risk classifier. PARSARC consists of the sequential assessment of tumour FGFR1 and PDGFRA protein levels, TP53 mutational status and expression levels of 229 pazopanib outcome, including a subgroup of long-term responders. Comparative analysis of gene expression profiles in tumours that exhibited intrinsic pazopanib resistance and in paired pre-treatment and post-progression samples from a long-term responding patient identified 3 pro-inflammatory cytokines as candidate drivers of clinical pazopanib resistance. These studies represent discovery data of putative biomarkers that identify subgroups of patients with distinct clinical phenotype that are independent of conventional histological classification. Further studies in independent tumour cohorts are required to provide validation for iTME-based prognostic biomarkers in early stage STS and for the PARSARC classifier as a predictive biomarker for the use of pazopanib in advanced STS.