Cancer progression is shaped not only by malignant cells, but also by their interactions with the surrounding tumour microenvironment. In colorectal cancer (CRC) and colorectal cancer liver metastases (CLM), these spatial interactions influence tumour growth, metastatic colonization, and clinical behaviour. However, conventional bulk and single-cell transcriptomic approaches either average complex tissues or disrupt tissue architecture. Spatial transcriptomics overcomes these limitations by preserving cells and transcripts within intact tissue sections.
The aim of this thesis was to investigate the spatial organization and heterogeneity of the tumour microenvironment in CRC and CLM using imaging-based spatial transcriptomics. In Paper I, in situ sequencing was used to map the fibrotic rim surrounding desmoplastic colorectal liver metastases. This study showed that the rim is not uniform, but consists of two spatially distinct zones: an outer liver-facing zone resembling a fibroinflammatory response, and an inner tumour-facing zone enriched in cancer-associated fibroblast features and extracellular matrix remodelling. We also observed angiogenesis within the rim.
Paper II compared the two main histopathological growth patterns in CLM: encapsulated/desmoplastic and replacement growth patterns. Replacement lesions were characterized by damaged hepatocytes and bifunctional hepatocyte-cholangiocytes populations at the tumour-liver interface, suggesting a microenvironment that may support tumour integration into the liver parenchyma. In contrast, encapsulated lesions displayed interferon-γ signalling in neoplastic and stromal cells, together with a zonated fibrotic capsule enriched in the outer part with activated hepatic stellate cells, cytotoxic T cells, angiogenesis signals; whereas in the inner part we observed cancer-associated fibroblasts, and extracellular matrix remodelling.
Paper III investigated why circulating tumour DNA (ctDNA) is detectable in some patients with localized CRC but low or absent in others. By integrating plasma sequencing, histopathology, immunohistochemistry, spatial transcriptomics, and in situ mutation detection, this study showed that ctDNA shedding is influenced by tissue architecture and immune niches, in addition to tumour size or stage. Tumours with detectable ctDNA were characterized by necrotic pseudoluminal structures, loss of epithelial barrier integrity, myeloid infiltration, and degenerated epithelial cells.
Paper IV introduced an analytical neighbourhood enrichment score for spatial omics data. This method replaces Monte Carlo label shuffling with an analytical formulation that estimates expected spatial neighbour interactions. Across multiple spatial omics datasets, the analytical score showed strong agreement with permutation-based approaches while improving computational speed, providing a scalable tool for studying spatial relationships in large datasets.
Together, this thesis demonstrates the value of spatial transcriptomics for dissecting tumour heterogeneity and tumour-microenvironment interactions in CRC and CLM. By preserving tissue architecture, spatial transcriptomics provides insight into metastatic growth patterns, ctDNA release, and cellular organization within tumours. Future studies will benefit from integrating spatial transcriptomics with additional omics layers and functional validation models, supporting tumour biology understanding and therapeutic and diagnostic strategies.