DNA mismatch repair deficiency (MMRD) is caused by the inactivation of the mismatch repair pathway, which corrects DNA replication-associated errors. MMRD results in an elevated rate of point mutations and indels, called hypermutation, which leads to a high burden of neoantigens and increased tumour immunogenicity. As a result, hypermutated cancers show the highest response rates to immune checkpoint blockade (ICB), a type of immunotherapy that has revolutionised the treatment landscape for multiple cancer types. Yet, durable responses are only observed in a subset of patients. Therefore, advancing our understanding of the molecular mechanisms underpinning variable responses to ICB is critical. Previous clinical and preclinical studies of MMRD tumours showed that the clonality rather than the burden of mutations determines the variable responses to ICB observed in the clinic. Given that the clonality of mutations is determined by the evolutionary dynamics and patterns of intra-tumour heterogeneity, it is paramount to elucidate the clonal dynamics of MMRD tumours. Moreover, genomics analyses of MMRD tumour evolution have relied on bulk whole-genome sequencing (WGS), which is limited to infer the clonal structure and dynamics of tumours due to the impossibility of assigning mutations to subclones and limited sensitivity for subclonal mutation detection. Thus, the evolutionary trajectories and the timing of the onset of the disease remain unknown for MMRD cancers. Recently, the advent of single-cell (WGS) methodologies allowed for the assignment of mutations to individual cells, making the reconstruction of single-cell lineage histories possible.

As a first contribution in my PhD, I have reconstructed the phylogenetic history of eight patient-derived organoids with MMRD at single-cell resolution using single-cell wholegenome sequencing. I found that the evolution of MMRD tumours is characterised by the accumulation of tens of thousands of clonal mutations during decades, which precedes a phase of rapid clonal expansion. In addition, I showed that MMRD tumours exhibit high levels of genomic heterogeneity, due to the increased rate of subclonal point mutations and indels they accumulate upon clonal expansion of the most recent common ancestor.

As a second contribution, I present a computational framework, termed ClonalSim, for inferring the timing of clonal expansions and the growth rate of tumours using single-cell phylogenies inferred from somatic mutations. I applied ClonalSim to two patient-derived organoids of MMRD cancers to infer the timing of clonal expansion and the fitness of cancer cells. First, I determined that MMRD deactivation happens during childhood. Secondly, I predicted the age of onset to be more than a decade or 7 years before diagnosis.

As a third contribution, I leveraged single-cell RNA sequencing atlases to study the degree to which cancer cells in human tumours dedifferentiate to a more primitive “stem-like” state, which is a known hallmark of cancer. Since quantitative evidence for this phenomenon is still lacking, I investigated the dedifferentiation stage in colorectal MMRD tumours and then expanded my analysis to 10.000 bulk-RNA sequencing data sets from pediatric and common adult tumours. My results indicate that cancer cells do not fully revert to an embryonic transcriptional state, as their genome-wide expression is closer to expression profiles of post-natal cells than those of gastrulation and embryonic cell types.