Breast cancer heterogeneity within the neoadjuvant window
Breast cancer affects ~55,000 women in the UK annually. Key to the reliable management of the disease is robust characterisation of an individual’s tumour, to best determine the most appropriate treatment and likely prognosis. High-throughput measurement of gene expression has driven recent developments in improved clinical patient stratification but difficulties in accuracy and sensitivity persist. Key to producing — and validating — predictive signatures of disease response and progression lies in utilising patient-derived material, to maximise a study’s relevance to clinical practice. Traditionally, measurements taken at diagnosis or surgery have been correlated with long-term outcomes, such as evidence of recurrence or metastasis. These post-operative studies often take place over the course of decades and are understandably restricted in terms of speed, efficiency and cost. In contrast, the pre-operative neoadjuvant setting allows expedited, short course studies that maximise the information available from a tumour whilst it is still in situ, allowing multiple characterisations of the tumour to be performed over time. The in situ nature of neoadjuvant study has allowed phenotypic and molecular characterisation to be performed in concert, allowing, for example, treatment response to be explained in terms of variation in gene expression. Supporting these neoadjuvant studies are routine biobanking operations, facilitating repeated tumour sampling and physiological characterisations. A happy side-effect of this indiscriminate sampling strategy is the generation of cohorts that do not fit the traditional neoadjuvant model and offer the potential to query alternative hypotheses. This thesis makes use of two of these cohorts to investigate each of (i) intra-tumour heterogeneity under conditions of no-treatment and (ii) long-term latent treatment resistance. Key results include demonstrating a biopsy method-specific effect on gene expression and revealing a role of epigenetic modification in endocrine treatment resistance, respectively. In addition, a more generally applicable methodology to illustrate and quantify the association of a continuous variable — such as gene expression — with outcome is described. Taken together, these parallel threads depict the emerging utility of the neoadjuvant setting in portraying difficult to model clinically relevant aspects of cancer treatment and response. The results will likely prove to guide clinical best practice as well as inform future studies, with the novel datasets generated allowing comparison, validation and further analysis.