A LIVING BIOBANK FOR
REVERSE TRANSLATIONAL STUDIES
Can we generate ex vivo cultures that reflect the features of HGSOC?
A BIOPSY PIPELINE TO GENERATE PRE-CLINICAL OVARIAN CANCER MODELS
Developing novel therapeutic interventions requires pre-clinical models that reflect human cancers. While judiciously selected cell lines provide tractable models, they under-represent the genetic heterogeneity exhibited by tumours and lack clinical annotation. By contrast, living biobanks have the potential to generate well-characterised and tractable models suitable for discovery research, drug screening, biomarker development and potentially clinical decision-making.
Facilitated by our location on The Christie Hospital site, we collect samples from patients with chemo-naïve and relapsed ovarian cancer, either as solid biopsies or ascites. Subsequently, we have established a workflow to generate ex vivo cultures of highly puriﬁed tumour cells, unfettered by contaminating, genetically normal stromal cells and the microenvironment, with extensive proliferative potential.
As proof of concept, we extensively analysed 15 ex vivo cultures derived from 12 patients, termed ovarian cancer models (OCMs), demonstrating that they possess the expected hallmarks of ovarian cancer. Single-cell shallow whole-genome sequencing and M-FISH revealed conspicuous karyotype heterogeneity, consistent with extensive chromosome instability.
Comparing OCMs revealed four karyotype classes: genomes dominated by whole-chromosome aneuploidies, rearranged chromosomes, monosomies or tetrasomies. Analysis of mitotic centrosome numbers and high-resolution time-lapse microscopy revealed highly heterogeneous mitoses, which are far more catastrophic compared with those observed in established cell lines, indicating that thus far we have underestimated the mitotic dysfunction in advanced human cancers.
For drug-sensitivity profiling, OCMs are transduced with a lentivirus expressing a GFP-tagged histone, enabling high-resolution microscopic analysis of population dynamics in response to therapy. EC50 values for cisplatin ranged ~7-fold across the cohort of 15 OCMs. Strikingly, those with the lowest cisplatin EC50 values were derived from patients who achieved a radiological response and CA125 reduction following platinum-based chemotherapy. By contrast, OCMs with high EC50 values originated from patients with progressive disease who did not demonstrate an improvement in CA125. This congruence between patient tumour responses and drug sensitivities ex vivo therefore demonstrates that the OCMs do indeed reﬂect the patients’ tumours.
Having validated that the OCMs recapitulate the features of ovarian cancers, we are continuing to build this living biobank. Thus, our research is supported by this foundation technology, which we use to probe underlying biological mechanisms and test novel therapeutic strategies in the context of ovarian cancer. We are also generating multi-omics datasets from the biobank, initially focusing on transcriptomics and proteomics, using RNAseq and SWATH-MS, respectively. With these datasets we aim to use our experience with supervised learning algorithms, combined with drug-sensitivity profiling of the OCMs, to identify and validate multi-omics signatures associated with drug sensitivity.
Nelson L, Tighe A, Golder A, et al. A living biobank of ovarian cancer ex vivo models reveals profound mitotic heterogeneity. Nat Commun. 2020; 11:822.
Pillay N, Tighe A, Nelson L, et al. DNA Replication Vulnerabilities Render Ovarian Cancer Cells Sensitive to Poly(ADP-Ribose) Glycohydrolase Inhibitors. Cancer Cell. 2019; 35:519–533.
Barnes, B. M., Morgan, R. D., Tighe, A, et al. Classification of ovarian cancer cell lines using transcriptional profiles defines the five major pathological subtypes. Available at: https://www.biorxiv.org/content/10.1101/2020.07.14.202457v1