Computational and Multi-omic Profiling of Resistant Tumour Cell Subclones in Oestrogen Receptor-Positive Breast Cancer

Mechanisms of resistance to aromatase inhibitors (AIs) vary between patients, and the rational use of the growing number of targeted agents for endocrine resistance will require new ways of identifying the specific resistance mechanisms operating within individual tumours. This project will address that challenge by analysing data and samples from the POETIC (Peri-Operative Endocrine Therapy – Individualising Care) study (4). POETIC was an open-label, multicentre, parallel-group, randomised, phase 3 trial (done in 130 UK hospitals) in which 4480 postmenopausal women aged 50 years or older with WHO performance status 0–1 and ER+, operable breast cancer were randomly assigned (2:1) to receive peri-operative AI treatment (POAI) (letrozole 2·5 mg per day orally or anastrozole 1 mg per day orally) for 14 days before and after surgery or no POAI (control). Adjuvant treatment was given as per UK standard local practice. In this cohort, we showed that KI67 levels, and changes Ki67 after POAI, were associated with time to recurrence (TTR) and overall survival (OS) (4). The Ki67 value was measured using a standardised visual scoring method in which stained and unstained invasive tumour nuclei were counted across at least 5 high-power fields (i.e., the most representing field); the Ki67 staining index was calculated as the proportion of stained invasive nuclei.
Intratumor heterogeneity of Ki67 expression after neoadjuvant endocrine therapy is thought to be associated with poor outcomes and treatment resistance but datasets and analytical approaches capable of resolving molecular differences between the resistant and sensitive subclones remain limited. The central biological question is whether persistent Ki67-positive cell populations represent resistant clonal populations that continue to survive under treatment pressure and ultimately drive micro-metastatic outgrowth and recurrent disease . This broader question can be broken down into three components.
 
First, are the ER+/Ki67+ persistent proliferating cells at 2 weeks molecularly distinct, either genetically or epigenetically, from the non-proliferating cells? In pilot project, multi-parameter fluorescence-activated cell (FACS) applied to 26 AI-treated FFPE POETIC samples collected at 2 weeks enabled separation of de novo AI-resistant from sensitive cell populations and, whole-genome sequencing showed a higher mutational burden in the AI-resistant populations at presentation (5). These early data suggest that systematic comparison of proliferative and non-proliferative tumour cell populations may identify putative drivers of resistance in individual tumours. This FACS-based approach provides an effective bridge between approaches that are lower-cost but labour-intensive methods that provide only limited resolution of tumour heterogeneity, such as needle microdissection, and high-cost approaches that provide a much richer view of heterogeneity, such as spatially resolved single cell sequencing.
As single-cell and spatial omics methods for FFPE tissues continue to evolve, the project will evaluate data from platforms such as cyclic immunofluorescence, 10x Visium, GeoMx Digital Spatial Profiling, and/or CosMx Spatial Molecular Imager. A major computational component of this project will be the development of approaches to integrate these data with digital pathology, including high-throughput image analysis and machine learning/artificial intelligence, in order to quantify morphological and biomarker-defined tumour subpopulations. Combining digital pathology with spatial and bulk multi-omics will allow us to interrogate resistant clones within their native microenvironment, and to connect patterns with molecular features at scale.
We will also examine whether resistant clones stratify according to HER2 status or intrinsic subtypes. A further key question is whether multi-omic markers identified in Ki67-persistent cells at 2 weeks are also detectable in recurrent disease, and whether these differ between locoregional and distant recurrence. POETIC provides a unique opportunity to address this. The protocol includes optional recurrence sample collection at the time of relapse. This will enable systematic comparison of multi-omic markers across matched primary tumours, locoregional recurrence, and distant recurrence, and will provide a clinically relevant framework for studying how early resistant-cell states relate to later disease progression.
 
Finally, this PhD project will investigate whether improved treatment strategies can be identified for AI–resistant clones. In collaboration with Translational Medicine and Bioscience colleagues at AstraZeneca, genomic and phenotypic features of resistant tumour cell populations will be compared with data from ER+ cell lines and patient-derived cancer models treated with investigational compounds. This comparative analysis will help to assess resistant-cell states align with responses to next-generation endocrine therapies or novel therapeutic combinations. The aim is to identify alternative therapeutic strategies, refine biomarkers for patient stratification, and define biological endpoints that could support adaptive, clinical trial designs.
 
By incorporating digital pathology, spatial omics, and bulk multi-omics, and clinically annotated trial data, this PhD project will generate a multi-layered view of endocrine resistance biology. It will be particularly suitable for students interested in computational biology, data science, machine learning, quantitative pathology, and translational genomics, while retaining strong links to clinical and experimental cancer research. Through partnership with AstraZeneca, the student will gain experience in both forward and reverse translation, and in how computational and molecular analysis can inform the design of next-generation precision oncology trials.