Integration of Omics Data to Explore Heterogeneity in Colorectal Cancer

Colorectal cancer (CRC) is the third most common cancer and the second leading cause of cancerrelated deaths worldwide. Recent advancements in high-throughput technologies have enhanced the understanding of this disease. RNA sequencing (RNA-seq) has allowed a better understanding of the transcriptome, helping to better describe the CRC molecular heterogeneity. The analysis of transcriptomic data has led to the development of the Consensus Molecular Subtypes (CMS) classification system, allowing the integration of both clinical and molecular characteristics of CRC. Small RNA sequencing (sRNA-seq) has shown great potential, identifying several microRNAs (miRNAs) involved in CRC development and progression, from precancerous lesions to metastasis. Due to their stability, miRNAs and other small non-coding RNAs can be detected in surrogate tissues, such as stool, urine, and blood, making them promising biomarkers. In parallel, in the last years, metagenomic sequencing of stool samples has shown that gut microbiome alterations are linked to CRC progression and treatment response. Integrating these omics data offers great opportunities to improve CRC classification and identify biomarkers for early detection and patient’s stratification. Furthermore, the possibility of detecting these molecular changes in non-invasive samples, such as stool, suggests promising new approaches for improving current CRC screening. RNA and small RNA sequencing were performed on tumor and adenoma tissues and their paired adjacent mucosa of 181 patients diagnosed with pre-cancerous lesions or CRC and collected in a cross-sectional study. To explore the landscape of sncRNAs in CRC, the expression levels of other non-coding RNAs were quantified after the alignment of human miRNome-unmapped reads and differential expression (DE) analyses were conducted to identify altered sncRNA patterns. Small RNA-Seq was performed on stool samples from a subset of these patients and colonoscopy-negative controls (n=87) to explore whether alterations could also be detected in this biospecimen. In the second part of the study, patients were classified into CMS subtypes using RNA-Seq data and each CMS was further characterized using a pan-cancer NGS assay. DE analyses across CMS subtypes were carried out to identify specific expression signatures within each subtype, and again, we explored whether CMS-specific alterations in sncRNA expression could also be found in stool. As the third step, the microbiome composition investigated in the same stool samples was explored for microbial abundances across different CMS. Finally, an integrative omics analysis was performed to integrate all features of interest identified from the previous omics to assess whether the classification of CMS subtypes could be improved and detected in stool samples. A total of 3,524 miRNAs and 15,509 other sncRNAs were identified. Comparing tumor tissue to adjacent mucosa, 410 DE sncRNAs were observed with a general overexpression of tRNAs and downregulation of snoRNAs. Twelve sncRNAs showed an increasing or decreasing trend of expression levels from pre-cancerous lesions to stage IV CRC, and these dysregulation patterns were also found in stool samples of cancer patients. DE analyses on CMS subtypes showed profound alterations in CMS2 and subtype-specific dysregulation patterns. Several of these features were found to be altered in a CMS-specific way also in the stool samples of the same patients. Metagenomics analysis showed that, compared to other subtypes, CMS1 shows a reduced species richness and has an increased abundance of microbial oral species, such as Gemella Morbillorum, and that some microbial species already known for their role in CRC, such a Fusobacterium Nucleatum, are more abundant in CMS4 patients. Integrative analysis revealed that the use of the expression levels of DE sncRNAs and species abundance levels in stool samples, alone or integrated, provide a good CMS subtype separation.