Chris De Witte

Cancer has a major impact on global health. Epithelial ovarian cancer (EOC) in particular is a disease with poor outcomes. In contrast to some other cancer types, survival rates of EOC have hardly improved over the last decades, despite efforts like more aggressive surgical approaches and the introduction of novel systemic therapies. The current five-year survival rate is 38%.

EOC is a heterogeneous disease on multiple levels. Firstly, EOC comprises a range of histological subtypes with distinct morphological and genetic features, of whom the most common as well as most aggressive subtype is high grade serous ovarian cancer (HGSC). Secondly, advancements in DNA sequencing techniques have revealed extensive genomic heterogeneity both between and within patients, especially in patients with HGSC. Interpatient genomic heterogeneity describes the genomic variation between tumors of different patients (with the same histological subtype) and intrapatient genomic heterogeneity describes differences in the genomic make-up of tumors from multiple sites of an individual patient.

Despite the observed heterogeneity, treatment for primary advanced EOC is fairly uniform and consists of debulking surgery and platinum-based chemotherapy. While most patients initially respond well to primary treatment, the majority of patients will experience recurrent disease and become resistant to chemotherapy. Preclinical model systems could aid in improving treatment stratification for individual patients by allowing treatment response assessment in vitro prior to administering the drugs in the clinic. Traditionally most preclinical research on EOC has been performed on 2D-cell cultures and xenografts, each with their own advantages and limitations. 2D-cell lines are fast-growing model systems that allow high throughput drug screening. However, they are also simplified tumor representations which lack cellular hierarchy. In contrast, xenografts (mouse models) are complex cancer models that allow in vivo drug response assessment but they require the use of animals and are not suitable for high throughput drug screening assays.

The organoid culture model was developed approximately 10 years ago and offers the opportunity to model patient-derived tumors in a three-dimensional manner, as the cells are expanded in a supportive extracellular matrix. The first organoid models were derived from healthy intestinal stem cells of mice in 2009, quickly followed by the establishment of human healthy intestinal organoids as well as organoids derived from patients with colorectal cancer. By now, patient-derived organoids (PDOs) have been established from multiple healthy and tumorous tissues by adapting the medium composition to specific tissue needs, including as of recently ovarian cancer.

In chapter 2 we describe the establishment of an EOC organoid biobank and its range of applications. A protocol is outlined, which enables the derivation and long-term expansion of PDOs of EOC. We established 56 organoids from 32 patients including all main subtypes of EOC. All samples were subjected to thorough characterization, as a proper patient-derived model system should faithfully represent the original tumor. PDOs recapitulated the histopathological and genomic features of the tumors they were derived from and were genomically stable over time. Gene expression analysis of PDOs revealed histopathological subtype-specific clustering. Further, PDOs were amenable to drug screening and in vitro drug responses were representative of subtype-specific clinical responses. Interestingly, drug response of organoids derived at different timepoints of an individual patient captured the transition from primary chemotherapy sensitive disease to recurrent chemotherapy resistant disease. Additionally, the transplantation of PDOs in mice enabled in vivo drug screening of patient-derived tumors. Chapter 2 also described the establishment and genetic modification of healthy fallopian tube (FT) and ovarian surface epithelium (OSE) organoids to mimic tumorigenesis. Overall, the results of this study show the wide potential of EOC PDOs for the research community, also confirmed by recent publications of other research groups on EOC PDOs.

In order for PDOs to obtain a role in clinical care, the correspondence between in vitro PDO drug response and in vivo patient drug response has to be established. In chapter 3 we compared PDO drug response and patient drug response to standard first line chemotherapy for seven organoids of five patients, to evaluate if PDOs can be used to predict clinical drug response. We established that PDOs recapitulated clinical drug response to first-line carboplatin and paclitaxel combination treatment for histological, biochemical and radiological outcomes. PDOs of patients with the best clinical response were most responsive in the drug screening assay, while patients with less responsive PDOs exhibited poorer clinical outcomes. Another requisite for incorporating PDOs in clinical care is a short turn-around time. We show in chapter 3, in line with results from other studies, that it is feasible to establish and screen PDOs within three weeks of tissue sampling.

While genomic heterogeneity of EOC has been studied extensively, the extent of functional heterogeneity is less known. We compared drug response to standard chemotherapeutics as well as targeted drugs for PDOs of multiple patients in chapter 3. PDOs exhibited both inter- and intrapatient drug response heterogeneity, which could partly be explained by genetic variation. The importance of our finding of extensive intrapatient drug response heterogeneity is that a single biopsy might not be sufficient for drug response prediction in individual patients.

To strengthen the correlation between in vitro PDO and in vivo patient drug response, a follow-up study is required. While we established a correlation between PDO drug response and clinical response in chapter 3, the sample size for this comparison was small and clinical drug response was limited to direct measures of response. Chapter 4 outlines a protocol for a prospective observational cohort study, in which we aim to establish the predictive value of PDOs for long-term clinical response, defined as progression free survival. We additionally aim to identify a cut-off for poor responders in the in vitro organoid drug screen assay, which could aid in identifying poor responders earlier. The outcomes of this study could aid in bridging the gap between the laboratory and the clinic for the use of PDOs. To improve the success rate of organoid establishment, which is currently limited for HGSC samples, we recommend an additional medium optimization step, prior to initiating the study described in chapter 4.

While WGS has been instrumental in characterizing PDOs in the previous chapters, we switch gears in chapter 5, where we focus on the analysis of WGS tumor data of a large cohort of patients with metastatic EOC. Treatment stratification is currently mainly based on histological subtype, disease stage, time to recurrent disease and response to prior platinum-based chemotherapy. We integrated WGS data of 132 solid tumor biopsies with clinical (treatment response) data to explore genome-informed treatment stratification opportunities. We show that unsupervised hierarchical clustering of genomic data identifies distinct clusters. Clusters were characterized by homologous recombination deficiency, genome stability and duplications and exhibited specific response rates and survival probabilities. Clustering allowed tumors to be classified beyond the traditional histopathological parameters. Moreover, an actionability analysis identified opportunities for genome-informed treatments, although in most cases it concerned off-label gene-drug interactions based on experimental evidence. Interestingly, the cluster with the worst survival rate contained the highest fraction of patients with an actionable target.

In conclusion, the work in this thesis resulted in the development of a novel preclinical PDO model system for EOC that is representative of the corresponding tumor lesions, is amenable to drug screening and captures both inter- and intrapatient genomic and functional heterogeneity. PDOs are valuable preclinical model systems that have the potential to provide insight in drug response for individual patients with EOC, complementary to genetic testing. A follow-up study which establishes the relationship between PDO drug response and long-term clinical response is warranted. Further, WGS can increase our insight in drug response and potentially improve treatment stratification for patients with metastatic cancer. By acknowledging the complex and heterogeneous nature of EOC and adopting a more individualized approach the outlookes for patients with EOC could ultimately be improved.