Investigating the clinicopathological spectrum and associated genetics of colorectal carcinoma in young (<60 years of age) patients in the Western Cape Province

Aldera, Alessandro

Investigating the clinicopathological spectrum and associated genetics of colorectal carcinoma in young (<60 years of age) patients in the Western Cape Province

Thesis / Dissertation

2025

Publisher

University of Cape Town

Department

Department of Pathology

Faculty

Faculty of Health Sciences

Abstract

The incidence of colorectal carcinoma (CRC) in young patients is rising in sub-Saharan Africa, and is set to become a major public health problem within the next decade. Despite this, there is a paucity of large-scale genomic studies in the subregion. To investigate driver genes, oncogenic signalling pathways and spectrum of pathogenic variants, we retrospectively identified 197 CRC cases over a 5 year period. Thirty-two mismatch repair deficient (dMMR) cases, without known germline variants, were investigated with amplicon-based panel next generation sequencing (NGS). Pathogenic or likely pathogenic variants were detected in the corresponding MMR gene in 14 of 18 (78%) MLH1/PMS2-deficient tumours, 5 of 8 (63%) MSH2/MSH6-deficient tumours, 1 of 4 (25%) tumours with isolated MSH6 loss, and 0 of 2 tumours with isolated PMS2 loss. Cases with a variant allele frequency suggesting a germline mutation were identified in MLH1 (eight), MSH2 (two) and MSH6 (one). NGS-based strategies for Lynch syndrome screening are advised to detect the broad spectrum of disease-causing MMR gene variants in our population. Resource constraints prohibit the rollout of universal MMR screening in sub-Saharan Africa. We sought to determine the performance of a deep learning model in our ethnically heterogeneous cohort. Our model yielded an AUROC of 0.91 (±0.02). Calibrating the classification threshold to 0.15, the overall sensitivity achieved in our cohort was optimised to 96% (95% CI 90-100) with a specificity of 60% (95% CI 52-82). This model could therefore be employed to accurately pre-screen for dMMR cases, thereby reducing the burden of downstream immunohistochemical and molecular testing in our resource limited setting. Whole exome sequencing was performed on a subset of the research cohort. Eighty-three cases were included in the analysis (77 MSS, 4 MSI, 2 POL). APC, TP53 and KRAS were among the most frequently mutated driver genes, although at a lower frequency than described in the literature. BRAF V600E mutations were absent. Although there were differences in the frequencies of mutations in the major driver genes, the frequencies of oncogenic pathway alterations were similar. FAT4 (26%) and TET2 (15%) have emerged as important novel driver genes in left-sided tumours, and potential therapeutic targets for further investigation.