Towards Personalised Medicine in Endometriosis: Creating Clinically Defined and Phenotype-Specific Models of Endometriosis

Download files Show statistical information Access & Terms of Use open access Copyright: Gunther, Kate Copyright: Gunther, Kate Altmetric Abstract Endometriosis is a chronic, inflammatory condition causing debilitating pain, fatigue and subfertility in an estimated 190 million people worldwide. Currently recognised as only one disease, endometriosis is defined by the presence of lesions which contain endometrium-like stroma and epithelium. However, emerging clinical and molecular evidence increasingly supports the hypothesis that endometriosis comprises multiple distinct subtypes, which may warrant the development of more targeted therapeutic options and personalisation of disease management. Progress in developing endometriosis treatment is currently limited by a lack of appropriate preclinical models for research. This thesis is composed of one published literature review and three independent research articles. Collectively, these works address the shortage of clinically relevant models of endometriosis by leveraging both historical data and tissues, as well as prospectively collected surgical specimens that were surplus to patient care. A comprehensive review was conducted on 122 publicly available omics datasets reported to represent endometriosis. This revealed that almost half of all publicly available datasets labelled as “endometriosis” did not represent true disease, instead containing only endometrial cells and tissues (59/122, 48.4%). Additionally, disease phenotype was unreported in a substantial portion of endometriotic datasets, and datasets contained very limited, if any, matched clinical data such as treatment response or recurrence status. These findings demonstrate the critical need for research to prioritise lesion-specific, clinically annotated datasets in order to ensure biological relevance and avoid potentially misleading conclusions. A recommendation to focus research on lesions was therefore strongly reinforced. To address this gap, historical tissues from patients enrolled in the Australian National Endometriosis Clinical and Scientific Trials Registry were used to generate a novel tissue microarray comprising 114 lesions, 56 adjacent normal tissue regions, and 32 utopic endometrial samples from 53 patients. Immunohistochemical analyses revealed subtype- and stage-specific patterns of biomarker expression, including higher Ki67 expression in advanced-stage disease and endometrioma lesions, and progesterone receptor loss associated with hormonal therapy and recurrent disease. Intrapatient heterogeneity in hormone receptor and proliferation marker expression highlighted the importance of multi-lesion sampling and demonstrated that single-lesion assessments may inadequately capture clinically relevant variability. These findings establish tissue microarrays as a reproducible ex vivo resource to investigate endometriosis heterogeneity and biomarker associations with clinical features. Complementing this, prospectively collected surgical specimens were used to establish patient-derived organoids from a range of endometriosis lesion subtypes. Overall, 78.6% of biospecimens successfully formed three-dimensional structures, and 53.6% remained viable after cryopreservation. Establishment rates varied by lesion phenotype, with endometrioma organoids exhibiting smaller size and greater reductions in progesterone receptor expression, and hormonal therapy at the time of surgery markedly reducing the success of organoid cultures. Organoids retained key morphological and molecular features of the original tissue, providing a scalable, in vitro platform to model patient-specific disease biology and evaluate therapeutic responses. Finally, the utility of these clinically annotated models was demonstrated in a drug repurposing study targeting ROR1. Transcriptomic analysis of 408 endometriosis samples and validation in 179 tissue microarray specimens confirmed ROR1 as a relevant candidate target. Approximately 9,000 compounds were screened in silico to identify 262 potential ROR1 inhibitors, from which three agents were prioritised for in vitro testing following clinical and patient input. Rimegepant, the most promising candidate, demonstrated dose-dependent reductions in organoid growth and viability across three patient-derived models, with interpatient heterogeneity reflecting lesion-specific biology. These findings highlight how integrated in silico, ex vivo and in vitro platforms can support translational research, inform targeted therapy development, and guide personalised treatment strategies for endometriosis. Collectively, this thesis advocates for the creation of diverse and clinically-relevant preclinical models of endometriosis which capture the clinical heterogeneity of the disease. It demonstrates how such models can be employed to identify candidate biomarkers and inform the development of targeted therapies, with the overarching aim of advancing clinical care for individuals affected by endometriosis.