Modelling the age-dependent dynamics of clonal haematopoiesis

Abstract

Somatic evolution, the process by which mutations accumulate and are subject to selection within cell populations, occurs across healthy tissues and contributes to both ageing and cancer development. In blood-producing stem cells, this process, known as clonal haematopoiesis (CH), can often be detected years before the onset of haematological cancer, raising the possibility that individual cancer risk could be inferred from a routine blood sample. While recent studies have begun to quantify CH dynamics -- for example, estimating expansion rates for common driver variants -- our understanding remains incomplete. This thesis addresses two key unanswered questions: how the selective landscape in blood changes across the human lifespan, and how CH can be distinguished from inherited variation in the context of genetic cancer risk screening. The large scale and broad age range of the UK Biobank (UKB) cohort offered an opportunity to investigate the long-term dynamics of CH without the need for longitudinal sampling. Somatic mutations were identified from blood-derived whole exome sequences of 420,000 cancer-free individuals aged 40-70 years. A mathematical model of clonal dynamics based on evolutionary theory was refined and adapted to enable cross-sectional estimation of mutation rates and fitness effects for CH driver variants across this age range. In 12 of the 15 most common known CH driver genes, including DNMT3A and TET2, the prevalence of clones carrying driver variants was observed to be substantially higher in younger individuals than would be expected under the assumption of constant growth. Furthermore, these clones were found to expand too slowly between ages 40 and 70 years to account for their observed cell fractions, suggesting that clonal growth was faster earlier in life and subsequently decelerated. For these driver genes, clonal expansion rates exhibited at least a 2- to 5-fold deceleration before the age of 40. In contrast, expansion of clones carrying SRSF2, SF3B1, and IDH2 variants showed no evidence of deceleration, consistent with previous findings. The evolutionary model was adapted to test two potential explanations for this deceleration. Firstly, variation in fitness between individuals with the same driver variant has the potential to cause apparent deceleration in cross-sectional analysis. After exploring a range of fitness distributions, the model was unable to reproduce the magnitude of the deceleration observed in UKB, suggesting that variation in fitness is not the primary cause of clonal deceleration. The second possibility was that competition between multiple expanding clones in the same individual causes clonal expansion rates to slow in later life. Theoretical analysis, modelling and simulation were used to estimate the genome-wide level of positive selection in blood required for widespread clonal competition to occur before age 40 years. The estimates were markedly higher than previously reported, and inconsistent with plausible levels of positive selection, suggesting that early-life deceleration of CH is unlikely to be driven by clonal competition. Pathogenic variants in TP53 are frequently identified in genetic risk screening, but whether a variant is inherited (germline) or acquired through clonal expansion (somatic) has important implications for cancer risk in both carriers and their relatives. In UKB, a classifier was developed to distinguish between germline and somatic TP53 variants using a combination of VAF distribution analysis and identification of shared haplotypes. 89% of carriers were classified confidently (70% somatic, 19% germline). The prevalence of TP53-driven CH was higher among current smokers and individuals with a previous cancer diagnosis. Survival analysis indicated that, among UKB participants with no previous history of cancer, the risk of cancer associated with high-VAF somatic TP53 variants was comparable to that associated with germline variants, but somatic risk was predominantly limited to haematological cancer.