Technology: Developing model systems to assess plasticity in cancer
Generation of patient-derived organoids
In the past, we have generated three-dimensional (3D) organoid cell cultures of mouse PDAC cells (Reichert et al., Nature Protocols 2013). With our expertise in 3D culture, we applied this technology to generate disease models directly from the patient. Over the past decade, organoid technology has reshaped biomedical and translational research in oncology and beyond. Importantly, tumor organoids maintain the genetic heterogeneity of the parenteral tumor, therefore these 3D structures represent a promising preclinical model system, particularly, for a heterogeneous cancer such as pancreatic cancer. Using this model system, we are able to make contributions to different areas in the field of cancer research (Renz et al., Cancer Cell 2018; Ruess et al., Nature Medicine 2018; Renz et al., Cancer Discovery 2018; Biederstädt et al., Gut 2020; Dantes et al. JCI Insight 2020; Feldmann et al., Gastroenterology 2021; Breunig et al., Cell Stem Cell 2021; Lesch et al., Nature Biomedical Engineering 2021; Zhang et al., Gastroenterology 2021; Krauss et al., Cancer Research 2021; Peschke et al., EMBO Mol Med 2022).
Studying tumor morphogenesis and metastasis using patient-derived organoids in in vivo systems
In order to study more complex biological processes, we applied patient-derived organoid (PDO) technology to, for example, orthotopic transplantation into pancreata of immunocompromised mice. Interestingly, the PDOX tumors resemble the histology of the primary patient-tumor to a remarkable degree and display metastatic dissemination (Dantes et al., JCI Insight 2020). More recently, we have developed the PDO-on-CAM model. The chick embryo chorioallantoic membrane (CAM) is a unique in vivo model that overcomes many limitations to study metastasis. Being naturally immunodeficient, the chick embryo accepts transplantation from various tissues and species without specific or nonspecific immune responses. Engrafted tumor cells invade through the basement membrane of the chorionic epithelium and into vascular structures in the underlying mesenchyme, thereby metastasizing to distant structures and organs of the developing chick embryo. Using quantitative PCR to detect human-specific Alu sequences present in xenografts, ectopic human tumor cells can be quantified in distant tissues. Importantly, PDAC cells seeded onto the CAM establish tumors that recapitulate the morphology of their parental counterparts. Now, with these assays to quantify metastatic dissemination in ovo, we are able to manipulate pancreatic cancer patient-derived organoids as well as murine PDAC organoids and test their capacity to metastasize. The scalability of the in ovo system allows performing medium-scale metastasis screening experiments using pharmacologic or genetic approaches.
Single-cell phenotyping of PDAC to study intratumoral heterogeneity and EMT plasticity
We employed digital holographic microscopy (DHM) as a novel tool for the high-throughput characterization of pancreatic cancer. DHM is based on the interference of an object and a reference beam. Once the light beam, which does not pass the object, interferes with the light beam passing the object, the interfering beams are recorded and quantitative phase information is reassembled from the recorded hologram. This phase contrast created by the differences in the refractive index at internal structures provides abundant intracellular information. Coupling the DHM to a microfluidic system allows us to analyze the cells not in an adherent state, but in a single-cell suspension in a high-throughput manner. By processing the acquired images with a pixel-based machine learning algorithm based on random forest classification, we are able to characterize the cells regarding cell type, EMT status, or malignancy – depending on our research question. Furthermore, we can divide pancreatic cancer PDOs with DHM similar to their respective transcriptomic subtyping into classical as well as quasi-mesenchymal subtypes. Being able to identify and differentiate all kinds of cell types and their respective cell stages in a label-free and high-throughput fashion opens up a broad variety of possible applications in clinical translation.
Next-generation organoids to identify distinct PDAC phenotypes
In order to visualize molecular differences that impact morphology, we modified established organoid culture conditions so that the cells are able to reshape the extracellular matrix (ECM) surrounding them. This we achieved by modifying the media conditions as well as by loosening the gels from the bottom of the culture dishes and generating floating gels. Interestingly, the composition of the ECM plays a critical role in this process. And indeed, distinct cell lines from distinct molecular clusters reproducibly generate complex 3D structures that allow inferring the molecular subtype. Tumor cells with an epithelial gene expression signature form thick branched structures whereas cells from the mesenchymal cluster form star-shaped organoids with spiky protrusions. Next, we applied deep convolutional neural networks for categorizing organoids into epithelial or mesenchymal morphologies based on imaging characteristics. By integrating this novel organoid technology with artificial intelligence, we are now able to set up large-scale therapeutic screens directed against cellular plasticity and specifically EMT and MET, based on phase-contrast images and computer-aided detection.