Dissecting the Heterogeneity of Intestinal Cancers and Predicting Subtype-specific Therapeutic Strategies

Abstract

Heterogeneity refers to the inherent diversity observed among tumor cells within a population. Diverse mechanisms contribute to the development of heterogeneity in intestinal malignancies arising from the large intestine and small intestine. Notably, genetic mutations, chromosomal abnormalities, and alterations in gene expression patterns are primary drivers. Epigenetic modifications, including DNA methylation and histone modifications, also play a crucial role in shaping heterogeneity. Additionally, microenvironmental factors, such as immune cell infiltration and tumor-stromal interactions, contribute to the heterogeneity observed within the tumor ecosystem. In this circumstance, heterogeneity is evident between cancers from different patients (inter-tumor heterogeneity) and within a single tumor (intra-tumor heterogeneity), posing challenges for clinical management and treatment. In the last few decades, molecular subtyping has been widely adopted as an effective approach for better diagnosis, prognostication, and treatment selection. Nonetheless, there are still many challenges persist and necessitate further attention and resolution.

The development of comprehensive molecular subtyping systems revolutionized our understanding of colorectal cancer (CRC) heterogeneity. In 2015, the consensus molecular subtypes (CMS) classification was introduced as a landmark classification by integrating transcriptomic-based classification systems to define four consensus molecular subtypes (CMS1-4) in CRC. These subtypes exhibit discrete biological characteristics, laying the foundation for clinical stratification and subtype-based targeted intervention. Specifically, CMS1 represents microsatellite instability immune tumors, CMS2 encompasses canonical tumors, CMS3 comprises tumors with metabolic alterations, and CMS4 characterizes tumors with a mesenchymal phenotype.

While the classification system was originally developed based on transcriptomic data from fresh-frozen (FF) tissues, a cost-effective method for accurately categorizing patient samples into CMS subtypes using expression profile derived from formalin-fixed paraffin-embedded (FFPE) samples is still limited, presenting a significant challenge in clinical practice for CRC subtyping. In Chapter 2, I introduced the CMSFFPE, an FFPE-based CMS classifier utilizing the expression levels of lowly degraded genes, demonstrating strong accuracy in predicting CMS for FFPE samples. I also validated the performance of this CMSFFPE gene classifier in FF tissues and demonstrated the prognostic relevance of CMS in colorectal cancer, thus enhancing patient prognostication and the advancement of precision therapeutic approaches.

Another challenge for CMS lies in the limitation of the use of bulk-tumor transcriptomic profiles, representing averaged expression signals of various types of cells within tumors. The tumor composition, including tumor tissue and tumor microenvironment (TME), for each subtype may remain heterogeneous. A growing appreciation of the role of the TME has driven the formation, development, relapse, and therapeutic sensitivity of cancer. Therefore, understanding the TME within the CMS subtypes is critical and remains to be investigated. In Chapter 3, I assessed the extent of tumor-infiltrating patterns for each CMS subtype and revealed two novel subtypes in CMS4, termed CMS4-TME+ and CMS4-TME-. CMS4-TME+ was characterized by upregulation in the infiltration of immune and stromal cells, and CMS4-TME- was distinguished by immune desert, stromal infiltration, and the poorest survival. Retrospective analysis of patients who received adjuvant chemotherapy demonstrated the CMS4-TME+ subgroup exhibited the most pronounced efficacy. In contrast, CMS4-TME+ might receive potential benefits in immunotherapy. Furthermore, to facilitate the application of the new classification system for CRC with enhanced efficiency and resource conservation, we estimated 360 spatial characteristics from conventional H&E pathological images using artificial intelligence (AI) techniques. Among them, the spatial diversity indices (SDIs), defined by the Shannon information entropy within the tumor and in the tumor proximal regions, captured the heterogeneity of tissue types in the TME. Finally, molecular subtype prediction from spatial characteristics was performed for prognostic and clinical management purposes.

In the majority of gastrointestinal cancers, the molecular classifications have been thoroughly studied. Nevertheless, the field of small intestinal adenocarcinoma (SIA) remains relatively unexplored. In Chapter 4, I constructed a molecular classification system for SIA, aiming to capture the molecular diversity within this poorly studied cancer type. Efforts are underway to develop classification systems that integrate genetic alterations, molecular pathways, and clinical features. More specifically, I first performed unsupervised clustering based on the gene expression profile of an inhouse SIA cohort (n=139), revealing four major subtypes (immunogenic, metabolism, immune suppression, and mesenchymal). Followed by mutation analysis and clinical characteristics. Finally, I conducted drug repurposing to predict therapeutic strategies specific to each subtype.

In this thesis, I developed a novel CMS classification model specifically for FFPE RNA samples; I integrated molecular and microenvironmental signatures to refine the CMS classification system; finally, I undertaken a comprehensive exploration of molecular subtypes within SIAs. Together, this thesis focusses on dissecting the heterogeneity of intestinal cancer, including CRC and SIA, and providing subtype-specific therapeutic insights.

Date of Award	19 Dec 2023
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Kui Ming CHAN (Supervisor), Xin Wang (External Co-Supervisor) & Praveen SETHUPATHY (External Co-Supervisor)